Masonary Arch Bridges

Masonry Arch Bridges
Background document D4.7
PRIORITY 6
SUSTAINABLE DEVELOPMENT
GLOBAL CHANGE & ECOSYSTEMS
INTEGRATED PROJECT
Sustainable Bridges SB-4.7 2007-11-30 2 (8)
This report is one of the deliverables from the Integrated Research Project Sustainable Bridges - Assessment for
Future Traffic Demands and Longer Lives funded by the European Commission within 6th Framework Pro-
gramme. The Project aims to help European railways to meet increasing transportation demands, which can only
be accommodated on the existing railway network by allowing the passage of heavier freight trains and faster
passenger trains. This requires that the existing bridges within the network have to be upgraded without causing
unnecessary disruption to the carriage of goods and passengers, and without compromising the safety and econ-
omy of the railways.
A consortium, consisting of 32 partners drawn from railway bridge owners, consultants, contractors, research
institutes and universities, has carried out the Project, which has a gross budget of more than 10 million Euros.
The European Commission has provided substantial funding, with the balancing funding has been coming from
the Project partners. Skanska Sverige AB has provided the overall co-ordination of the Project, whilst Lule Tech-
nical University has undertaken the scientific leadership.
The Project has developed improved procedures and methods for inspection, testing, monitoring and condition
assessment, of railway bridges. Furthermore, it has developed advanced methodologies for assessing the safe
carrying capacity of bridges and better engineering solutions for repair and strengthening of bridges that are found
to be in need of attention.
The authors of this report have used their best endeavours to ensure that the information presented here is of the
highest quality. However, no liability can be accepted by the authors for any loss caused by its use.
Copyright Authors 2007.
Figure on the front page: Output from FEM analysis - principal stresses and displacement vector.
Project acronym: Sustainable Bridges

Project full title: Sustainable Bridges Assessment for Future Traffic Demands and Longer Lives
Contract number: TIP3-CT-2003-001653
Project start and end date: 2003-12-01 -- 2007-11-30 Duration 48 months
Document number: Deliverable D4.7 Abbreviation SB-4.7
Author/s: C. Melbourne, J. Wang, A. Tomor, USAL
G. Holm, M. Smith, P.E. Bengtsson, SGI
J. Bien, T. Kaminski, P. Rawa, WUT
J. R. Casas, P. Roca, C. Molins, UPC
Date of original release: 2007-11-30
Revision date:
Project co-funded by the European Commission

within the Sixth Framework Programme (2002-2006)
Dissemination Level
PU Public X
PP Restricted to other programme participants (including the Commission Services)
RE Restricted to a group specified by the consortium (including the Commission Services)
CO Confidential, only for members of the consortium (including the Commission Services)
Summary
This background document contains the basic and full information summarized in chapter 8
of the Guideline SB-Resist (2007): Guideline for load and resistance assessment of railway
bridges, with reference to the assessment of masonry arch bridges.
Recent survey carried out by WP2 showed that 40% of existing rail network bridges
in Europe are masonry arches. More than 60% of those bridges are over 100 years
old and still carry ever increasing levels of loading and increasing volumes of traffic.
Due to the constantly increasing weight of rail traffic there is increasing demand for
better understanding of their life expectancy and fatigue limits. It is imperative that
the bridge stock is not adversely affected by these changes in the loading regime and
that appropriate assessment, modelling, repair and strengthening techniques are
available.
Over the past 10 years there has been an extensive programme of research which
considered some aspects of masonry arch behaviour. A number of small and large
scale tests have been carried out on masonry arches, most of them have been under
static (monotonic) loading and considered mainly the arch ring itself. There has been
also very little work done on investigating the long-term effect of traffic (cyclic) load-
ing and the effect of deteriorated masonry on the fatigue life of the bridge.
Most tests focussed on the behaviour of the arch ring itself and little work has been
done to investigate how the loads from the train traffic are transferred onto the arch
ring and how much the soil-structure interaction may contribute towards the capacity
of the bridge.
Additionally, most of the available methods of assessment are deterministic. The
proper treatment of uncertainty in the assessment of metal or concrete bridge can
save it from strengthening or replacement. Many practical cases have shown the po-
tentiality of reliability-based assessment methods for concrete and steel bridges.
However, their application to masonry bridges has been almost negligible, despite
some experiences have shown the potentiality of the use of such methods in the
case of masonry arch bridges.
The background document is an attempt to address and solve the above questions.
To this end, the document is divided in the following parts:
D4.7.1 Structural assessment of masonry arch bridges

D4.7.2 Numerical analyses of load distribution and deflections in railway bridge tran-
sition zones due to passing trains
D4.7.3 Methods of analysis of damaged masonry arch bridges
D4.7.4 Potentiality of probabilistic methods in the assessment of masonry arches
Part D4.7.1 includes the analysis of failure mechanism in multi-ring brickwork arches
as well as the fatigue performance of multi-ring brickwork arches and the soil-
structure interaction.
A series of tests have been carried out at the University of Salford on large-scale
multi-ring brickwork arches (3m and 5m span, two and three-ring) under long term
cyclic loading. Tests showed that although the classical mode of failure of arches un-
der static loading is generally seen as the four-hinge-mechanism, all arches within
the present test series failed by ring separation under fatigue loading. The fatigue
loading reduced the load capacity to as low as 37% of the static load capacity for
two-ring arches and 57% for three-ring arches and was a function of ring separation.
Old masonry arch bridges often have suffered serious material deterioration as a
consequence of environmental effects, cyclic/fatigue effect, water penetration, etc
which can seriously reduce the load bearing capacity of the bridge and change its
critical modes of failure. Of particular concern is mortar loss from between the rings
of a multi-ring brickwork arch, resulting in ring separation with the consequential loss
of carrying capacity. A series of 5m span arches built of weak bricks were tested to
represent aged, poor quality masonry. Weak brick arches which have been tested so
far have shown reduced fatigue capacity of the arch by ca. 20% but did not greatly
change the mode of failure compared to strong brick arches.
Based on the test results, a model for an interactive S-N (ISN) curve has been sug-
gested. The ISN curve could be developed for each bridge, based upon standard
curves and modified according to the engineers assessment of the structural condi-
tion. It would allow qualitative assessment of the residual life and critical modes of
failures. It would also help decide the appropriate repair technique and quantify the
new residual life for a given load regime. The anticipated ISN curve based on the test
results is also presented in the study and the effect of material deterioration have
also been included.
Results from three large-scale bridge tests have demonstrated the importance of the
type of backfill and support conditions and have produced high quality data that can
be used to validate numerical models. They also indicated that the test rig performed
as designed (giving effectively plane strain conditions); inclusion of large windows
along one side of the test chamber permitted the acquisition of good particle image
velocimetry (PIV) data which is being used to help better understand the nature of
the soil-arch interaction.
Part D4.7.2 deals with the soil-structure interaction and the dynamic response of em-
bankment and backfill to railway loading.
In order to model and assess the load capacity of arch bridges accurately, it is nec-
essary to gain clear understanding of how much of the axle loads pass through the
backfill and how the arch is seeing the passing traffic. The influence of higher axle
loads and higher train speed on the properties/behaviour of the backfill and em-
bankment material will be investigated by numerical modeling. The long-term effects
of increased train loads and increased train speed on the stress limits of the structure
will be investigated in order to model the residual life of the bridge.
The SGI has studied the dynamic soil-loading problem using typical embankment
parameters and representing the bridge structures by rigid boundaries. The numeri-
cal modelling technique FLAC3D was used. The scope of this work included perform-
ing numerical analyses to evaluate the load distribution and deflections in bridge
transition zones due to the passing of trains. No field or laboratory tests were per-
formed. Numerical analyses consisted of three-dimensional, dynamic analyses. Us-
ing a bridge representative of a typical European concrete bridge, a numerical para-
metric study was performed to study the effects of 1.) the train load, 2.) the train ve-
locity, and 3.) the stiffness of the ballast, sub-ballast and the embankment fill. Key
findings from the work include:
The greatest increases in vertical and horizontal stresses are concentrated
within the upper 1.0 to 1.2m of the system. The additional vertical and horizon-
tal stresses caused by the passing train decrease with depth.
As train axle loads increase, the vertical and horizontal stresses, and vertical
deflections beneath the train increase.
As train velocity increases, the horizontal stresses and vertical deflections in-
crease, however, they do not increase linearly with velocity. As the velocity of
the train approaches the Rayleigh wave velocity of the embankment system,
significant horizontal stresses and vertical deflections develop.
Vertical pressures are affected more by train axle load, and the horizontal
pressures by train velocity.
As the stiffness of the ballast and/or sub-ballast materials increase, net hori-
zontal stresses exerted on the bridge decrease. Horizontal stresses in the
embankment fill are relatively small, but increase when a soft fill is used.
When a soft embankment fill is used, vertical deflections increase substan-
tially, and these deflections are observed to occur 3m behind the rain, as if
there is a stern wave behind the train.
Numerical analyses were also performed using a bridge geometry representative of
European masonry arch bridges. It can be assumed that the factors that affect the
stress distribution and deflections for the concrete bridge structure (as described
above) will have the same effect for the masonry arch bridge structure. However, the
geometry of the arch bridge is different than the concrete bridge in two significant
ways: (1) for masonry arch bridges there is a gradual change in stiffness, while for a
concrete bridge the change is more abrupt, and (2) for the masonry arch structure,
the ballast and subballast are confined within a relatively small volume. Due to the
latter factor, additional vertical and horizontal stresses induced by the train are dis-
tributed immediately to the arch structure. Since there is less soil above the crown of
the arch to dissipate stresses, the crown of the arch will experience the greatest in-
crease in vertical stress under the passing of a train.
The above work was verified against data from an instrumented test embankment
published by the Finnish Railway administration. These data referred to an embank-
ment which did not include a bridge. The results give some insight into the behaviour
in bridge transition zones. Due to the lack of funding, field tests and/or laboratory
tests in a real bridge close to an embankment were not developed to verify the nu-
merical analyses.
Part D4.7.3 is focussed on the effect of damages on carrying capacity.

Most masonry arch bridges have existing damages. Some of these bridges may not
pass first-line routine assessment and need to be assessed using more advanced
techniques. There are several analytical, computer based and semi-empirical meth-
ods available which may be used for assessing masonry aches. However not all of
them are suitable for modelling all types of existing damages in the bridge. The pre-
sent part of the background document gives an overview of the masonry arch dam-
ages and guidance on which assessment methods are suitable for various types of
damages and on their limitations for use.
Finally, part D4.7.4 reports a feasibility study of the application of probabilistic tech-
niques to the assessment of existing masonry arches.
Probabilistic assessment methods are not generally used for masonry arch bridges.
One of the reasons of its limited application is the difficulty of defining reliable failure
criteria for these types of structures. Experimental tests have shown that failure of an
arch is generally of a global nature rather than by failure of individual components.
There is also a lack of accurate theoretical models to idealise the behaviour of ma-
sonry arch bridges as a system.
Part D4.7.4 of the background document presents a complete methodology for the
probabilistic assessment of masonry arches at the serviceability and ultimate limit
states. The document explains the definition of the different failure modes and corre-
sponding limit state functions that may occur depending on the type of masonry con-
struction (mainly single-ring and multi-ring). Also the introduction of the possibility of
the fatigue failure of masonry arch bridges and the proposal of new assessment
methods based at the serviceability level and not only at the ultimate level of the 4
hinge-mechanism is a promising initiative in the field. As a consequence, the present
document has deeply investigated the potentiality of a probabilistic approach in the
estimation of fatigue safety and remaining service life of masonry arch bridges. As a
result, the need of simple numerical models to perform fast simulation trials has been
enhanced. Thus, the influence of important features in the bridge response, as the
arch-backfill interaction, the modelling of multi-ring arches and the bridge skewness
in the development of simplified theoretical models at the ultimate and serviceability
limit states have been worked out too.
The analysis developed in D4.7.4 has verified the results obtained in laboratory tests
regarding the influence of skewness in the failure load by 4-hinge mechanism. In fact,
it has been confirmed that the influence of a skew angle less than 22.5 is almost
negligible. The study has confirmed that only with advanced 3D FEM models is pos-
sible to correctly model the behaviour up to failure of masonry arches with skew big-
ger than 22.5. It has been also concluded that the worst load position for the arch
can be deduced using a simple model.
In summary, a methodology is presented for the fatigue and serviceability assess-
ment of masonry arches, and probabilistic S-N curves for the behaviour of masonry
under fatigue are developed based on the few experimental data available. In the
case of masonry under compression, a fatigue equation with various levels of prob-
ability of failure is also proposed that may be used for deterministic assessments. In
the case of masonry under shear a fatigue equation with a 50 % confidence level is
also proposed. The last can be of relevant application to the case of multi-ring
arches. Of course, these are only preliminary models and these fatigue models have
to be updated as more experimental data becomes available from future laboratory
and full-scale tests. Finally, the examples presented show how a probabilistic as-
sessment of masonry arch bridges which were rated unsafe using standard linear
models, indicated sufficiently high level of safety margins when assessed together
with appropriate non-linear models.
In the course of the research it became apparent that the current methods of as-
sessment lack a single multi-level (initial, intermediate, enhanced) approach that in-
cludes all the current methods of assessment/analysis and gives clear guidance on
the philosophy that governs the determination of the safe working loads and ultimate
load carrying capacity.
The SMART (Sustainable Masonry Arch Resistance Technique) assessment
method, developed as a result of the present research, achieves this. There are 7
steps in the method which are summarized in the following table:
INITIAL INTERMEDIATE ENHANCED

1) GEOMETRY & Basic dimensions Full geometrical Full geometrical
CONSTRUCTION for MEXE as- survey survey
sessment
2) LOADING Determined from UIC loading or UIC loading or
nomogram/charts equivalent equivalent
3) MATERIALS Identified and Field tests and Extensive field
condition as- samples tests and sam-
sessed. pling
4) ANALYSIS No analysis re- Mechanism Mechanism
quired for MEXE method plus con- method and
assessment sideration of other FE/DE methods
modes of failure including sophisti-
(eg ring separa- cated material and
tion) soil material prop-
erties
Statically determi-
nate analysis to
establish stresses
5) ULS N/A Check that the Check that the
failure load is failure load is
greater than the greater than the
factored loading factored loading
6)PLS Determined by the Check that the Check that the
application of fac- working loads do working loads do
tors to the initial not induce not induce
permitted axle stresses greater stresses that are
loading than permissible greater than the
stresses permissible
stresses if the
do then probabilis-
INITIAL INTERMEDIATE ENHANCED

tic techniques may
be used to con-
sider risk or if S-N
type curves are
available for the
critical stress pa-
rameters then this
can be considered
(see residual life)
7) RESIDUAL Not determined Not determined Determination
LIFE possible using
enhanced material
property parame-
ters (S-N type
curves) and/or
probabilistic tech-
niques
The Ultimate Limit State (ULS) is defined as the condition at which a collapse
mechanism forms in the structure or its supports. The Permissible Limit State (PLS)
is defined as the limit at which there is a loss of structural integrity which will meas-
urably affect the ability of the bridge to carry its working loads for the expected life of
the bridge.
The writing of the different parts as well as the review has been carried out according
to what is presented in the following table:
Part Responsible Contributors Reviewer

D4.7.1 Clive Melbourne, Adrin Tommor, USAL Joan Casas, Pere
USAL Roca, UPC
D4.7.2 Gran Holm, SGI Miriam Smith, Per Evert Clive Melbourne,
Bengtsson, SGI USAL
D4.7.3 Jan Bien, WUT Tomasz Kaminski, Pawel Rawa, Clive Melbourne,
WUT USAL
D4.7.4 Joan Casas, UPC Pere Roca, Climent Molins, UPC Tomasz Kaminski,
WUT
Structural Assessment of Masonry Arch Bridges
Background document D4.7.1
PRIORITY 6
INTEGRATED PROJECT
Sustainable Bridges SB-4.7.1 2007-11-30 2 (72)
Future Traffic Demands and Longer Lives funded by the European Commission within 6th Framework
Programme. The Project aims to help European railways to meet increasing transportation demands, which can
only be accommodated on the existing railway network by allowing the passage of heavier freight trains and faster
unnecessary disruption to the carriage of goods and passengers, and without compromising the safety and
economy of the railways.
the Project partners. Skanska Sverige AB has provided the overall co-ordination of the Project, whilst Lule
Technical University has undertaken the scientific leadership.

Document number: Deliverable D4.7.1 Abbreviation SB-4.7.1
Author/s: C. Melbourne, J. Wang, A. Tomor, USAL
Revision date:

Dissemination Level
PU Public X
Table of Contents
Summary .................................................................................................................... 5
1. Masonry arch behaviour under long-term cyclic loading ........................................ 7
1.1 Background.................................................................................................... 7
1.2 Objectives ...................................................................................................... 8
1.3 Arch construction ........................................................................................... 8
1.3.1 Arch dimensions ................................................................................. 9
1.3.2 Material properties .............................................................................. 9
1.3.3 Loading arrangement........................................................................ 10
1.3.4 Instrumentation ................................................................................. 11
1.4 Static loading tests....................................................................................... 12
1.4.1 Strong brick arches ........................................................................... 12
1.4.2 Weak brick arches ............................................................................ 19
1.5 Cyclic loading tests ...................................................................................... 21
1.5.1 Strong brick arches ........................................................................... 21
1.5.2 Weak brick arches ............................................................................ 28
1.6 Summary ..................................................................................................... 32
1.6.1 Static loading summary..................................................................... 32
1.6.2 Cyclic loading summary .................................................................... 34
1.6.3 SN curves ......................................................................................... 35
2. Soil-structure interaction in masonry arch bridges................................................ 37
2.1 Background.................................................................................................. 37
2.2 Objectives .................................................................................................... 37
2.3 Experimental programme............................................................................. 38
2.3.1 Test rig.............................................................................................. 38
2.3.2 Bridge geometry and material properties .......................................... 39
2.3.3 Bridge construction ........................................................................... 41
2.3.4 Bridge instrumentation ...................................................................... 45
2.3.5 Test Procedure and loading arrangement......................................... 48
2.3.6 Test results ....................................................................................... 49
2.4 Finite element modelling .............................................................................. 54
2.4.1 Finite element modelling approach ................................................... 54
2.4.2 Case study - flexible and smooth strip footings on stratum of clay.... 58
2.4.3 Finite element models....................................................................... 60

2.4.4 Finite element results and discussion ............................................... 63
2.5 Conclusions ................................................................................................. 68
3. Acknowledgement ................................................................................................ 69
4. References ........................................................................................................... 70
Summary
The Sustainable Bridges project survey has shown that 40% of the existing rail
network bridges in Europe are masonry arch bridges and that more than 60% of them
are over 100 years old. It is imperative that the bridge stock is not adversely affected
by changes in the loading regime and bridge condition and that appropriate
assessment, modelling, repair and strengthening techniques are available.
Research at the University of Salford has focused on two main issues relating to
masonry arch bridges that can significantly influence the bridges capacity and
service life:
fatigue performance of multi-ring brickwork arches
soil-structure interaction in masonry arch bridges.
Fatigue performance of multi-ring brickwork arches

A series of laboratory tests have been carried out at the University of Salford on
large-scale multi-ring brickwork arches under long term cyclic loading. Although four-
hinge-mechanism is regarded as the classical failure mode for masonry arches, all
arches within the test series under long-term fatigue loading failed by ring separation.
Although fatigue limit is currently estimated to be around 50% of the static capacity,
laboratory tests have shown it to be low as 40% of the static capacity. Based on the
test results, a model for interactive SN curves has been proposed for masonry
arches and masonry qualities that allow qualitative assessment of the residual
service life under a given loading regime.
Soil-structure interaction in masonry arch bridges

It is well established that soil-arch interaction has a significant influence on the load
carrying capacity. In order to model and assess the load capacity of arch bridges
accurately, it is necessary to gain clear understanding of how much of the axle loads
pass through the backfill and how the arch is seeing the passing traffic.
Results from three large-scale bridge tests have demonstrated the importance of the
type of backfill and support conditions and have produced high quality data that can
be used to validate numerical models. They also indicated that the test rig performed
as designed (giving effectively plane strain conditions); inclusion of large windows
along one side of the test chamber permitted the acquisition of good particle image
velocimetry (PIV) data which is being used to help better understand the nature of
the soil-arch interaction.
Research described within the current project and by others has confirmed that
masonry arch bridges behave in a complex way and that simple semi-empirical
assessment methods are no longer appropriate to justify an increase in load carrying
capacity. Consequently, a new holistic masonry arch bridge assessment strategy
(Sustainable Masonry Arch Resistance Technique - SMART) has been proposed that
incorporates the currently available assessment methods and takes recent research
and findings into account. The new Sustainable Masonry Arch Resistance Technique
(SMART) sets out a strategy by which the residual life and permissible limit of
masonry arch bridges can eventually be determined with the help of a set of SN
curves (see D4.2 Chapter 8 for details). It can help prioritise conflicting maintenance
demands on limited budgets and improve asset management. The SMART
assessment method is currently in its infantry and large volumes of material test data
are required for further development of the technique.
1. Masonry arch behaviour under long-term cyclic loading

1.1 Background
Railway bridges carry ever increasing levels and volumes of traffic. Experience in the
past 30 to 40 years seems to suggest that repeated application of heavy traffic loads
could accelerate the deterioration of masonry arch bridges. Most experimental work
to date has however been under monotonic loading conditions and there is a fear
that under cyclic loading these phenomena may be aggravated by the incremental
deterioration of the brickwork.
Failure of masonry arch bridges is typically associated with three modes:
formation of a hinge mechanism
snap-through failure prior to the full formation of hinges
crushing failure.
These failure modes generally relate to single span, single ring square arches without
taking the effects of soil-structure interaction, possible abutment movements, mortar
washout, long-term fatigue deterioration, etc. into account. In addition to the classical
failure modes, masonry arches may also fail by a number of alternative failure
mechanisms, such as ring separation, sliding, abutment movement, etc. (see Figure
1.1).
a) Four hinge mechanism c) Sliding
b) Ring separation d) Abutment movement

Figure 1.1: Masonry arch failure modes
Railway bridges in Europe almost invariably contain multiple arch rings. These rings
are either simply connected by uninterrupted mortar joints or crossed by headers
which provide an interlocking effect and greatly increases the shear strength of the
connection between rings (see Figure 1.2). The type of arch construction shows
certain regional consistencies around Europe, for example arch bridges in Britain are
generally built without headers, while in Southern Europe arches with interlocking
headers are more common.
The shear resistance of arch rings is also influenced by the quality of mortar and
brick units. Aged arch bridges often suffer from weak shear connections from washed
out mortar joints and/or weak interlocking bricks that can result in unexpected ring
separation and may significantly reduce the load carrying capacity of the bridge.
The currently described test series was designed to represent multi-ring arches
without headers under long-term traffic loading to investigate their likely modes of
failure and fatigue capacity.
a) Single-ring c)multi-ring b) multi-ring with headers

Figure 1.2: Different degrees of shear connections
1.2 Objectives
Objectives of the test series were to investigate the behaviour, possible modes of
failure, capacity and permissible limit state of multi-ring masonry arches under long-
term fatigue loading.
1.3 Arch construction

The arches were constructed on reinforced concrete abutments bolted into the
reinforced strong floor. The abutments were additionally held together with high
strength steel bars for the 5m arches. The springing surface was inclined at 44o
angle. The centering consisted of two curved steel girders and timber shuttering
which were removed 21 days after the construction of the arch. Constriction of the
arches took 1 day for the 3m span arches and 2 days for the 5m span arches.
Testing took place at least 28 days after construction.
1.3.1 Arch dimensions

Dimensions for the 3m and 5m span arches are listed in Table 1-1. The number of
brick units varied slightly in the arch rings and are listed in Table 1-2 separately for
the 2-ring and 3-ring arches.
Table 1-1. Arch dimensions
3 m arch 5 m arch
Span (mm) 3000 5000
Rise (mm) 750 1250
Ring thickness (mm) 215 330
Arch width (mm) 445 675
Number of rings 2 3
Dead load (kN) 2 x 10 2 x 22.5
Span : rise 4 :1
Shape Semi-circular
Table 1-2. Max. number of bricks in arch rings

3m arch 5m arch
Intrados Extrados Intrados Middle Extrados
(2-rings) (3-rings)
A 47 50 M 77 80 83
C 48 50 O 77 80 83
E 47 50 S 77 80 82
G 47 49 T 75 80 81
U 75 80 82
1.3.2 Material properties

Two types of 215 x 102.5 x 65mm bricks were used for construction: strong class A
engineering bricks to represent high quality brickwork and weak Britley Olde English
bricks to represent aged brickwork which is commonly found in the railway network.
For all tests 1:2:9 (cement:lime:sand) mortar was used. Material properties for the
bricks, mortar and triplets are shown in Table 1-3. Although the class A engineering
bricks were significantly stronger than the Britley Olde English bricks, their masonry
strength were not greatly different. Masonry strength has been shown by Lange
(1993) to be primarily dependent on the brick surface texture rather than the brick
strength.
Table 1-3. Material properties

Compression Density
[N/mm2] [kN/m3]
Brick (Strong) 154 23.7
Brick (Weak) 18.9 16.2
Mortar 1.86 15.5
Masonry triplet (Strong) 18.2 21.8
Masonry triplet (Weak) 11.7 16.9
Masonry prism (Strong) 24.5 22
Masonry prism (Weak) 9.1 16
1.3.3 Loading arrangement

In order to represent the weight of the typical backfill on arches, dead loads were
applied at the and points of the arches either by steel weights or by hydraulic
jacks (see Table 1-1 for dead loads, Figure 1.3 and Figure 1.4 for loading
arrangement). Live load was applied at the point for static tests by 1kN increments
and at the and points for cyclic tests. Cyclic loading was applied at 2Hz
frequency to represent the flow of traffic at ca. 25-30 miles/hours speed over the
bridge. Cyclic loading was applied for 106 cycles at each load level, starting from a
relatively small load. If after 106 cycles no damage or deterioration was observed, the
load was increased by 2kN and the process repeated until failure occurred.
Hydraulic jack
Figure 1.3: Loading frame
Figure 1.4: Hydraulic jacks

1.3.4 Instrumentation
Horizontal and vertical deflection was measured at the and span using LDVTs
(see Figure 1.5).
H () H ()
V () V ()
Figure 1.5: Deflection gauge positions
Vibrating wire gauges (GAUGE TYPE T/S/R, 139mm gauge length) were used to
monitor surface strain with a range of 0.5 - 3000 microstrains and were attached to
the masonry surface with an epoxy resin adhesive.
In addition to the standard test methods acoustic emission (AE) technique was
applied during the tests to gain deeper understanding of the fracture process. For
acoustic emission (AE) monitoring a Physical Acoustics DiSP system was used with
four channel PCI DSP4 boards and filter set to 20-200kHz. Preamplifiers were IL40D
with 20-100kHz frequency. The sensors were Physical Acoustics R6D resonant
piezoelectric sensors with a response range of 40-100kHz. The sensors were
selected based on advice and past experience as a good compromise of detection /
distance and avoiding noise from the loading rig. The filter bandwidth was broader
than the preamplifier and sensors to ensure that signals were not distorted. AE WIN
software was used to process data and in addition to the AE hit recordings load and
deflection were also measured every time emission was received. Up to eight AE
sensors were attached to the brick surface using a thin layer of hot-melt glue. This
technique had been tested to provide good coupling for transmitting AE signals.
1.4 Static loading tests
1.4.1 Strong brick arches
1.4.1.1 Arch A (3m span)

Arch A was loaded gradually up to failure. Four hinges opened gradually as shown in
Table 1-4, Figure 1.6 and Figure 1.7. The arch became unstable after opening of the
4th hinge near the span and failed by four-hinge mechanism under 29kN. Live load
was removed when the arch became unstable.
Live load was subsequently re-applied to identify any residual strength the damaged
arch may have in the presence of four radial cracks. Four-hinge mechanism re-
opened under 14kN live load - at 48% of the load capacity of the healthy arch.
Table 1-4. Arch A: Crack history
Live load Crack location
Hinge Type
(kN) Intrados Extrados
1st 6.5 14/15 15/16 Radial crack
2nd 14 0/1 0/1 Radial crack
rd
3 ? (23) 47/ 50/ Radial crack
th
4 23 27/28 29/30 Radial crack
DL + LL DL
23kN
6.5kN
14kN (?)
Figure 1.6: Arch A: failure mechanism
Figure 1.7: Arch A: Failure mechanism

Horizontal and vertical deflection was measured under the and span. No
changes in deflection were recorded during the opening of the 1st and 2nd hinges (see
in Figure 1.8). First change in deflection was shown from around 21kN, just before
the 3rd hinge was observed. Deflection sensors were removed from the arch at 23kN
to avoid damage.
Figure 1.8: Arch A: Load-Deflection

1.4.1.2 Arch G (3m span)

Arch G was also loaded gradually up to failure. Radial hinges ware observed as
shown in Table 1-5, Figure 1.9 and Figure 1.10. The arch became unstable after
opening of the 4th hinge near the span and failed by four-hinge mechanism under
28kN.
Table 1-5. Arch G: Crack history
Hinge Type
1st 22 14/15 16/17 Radial crack
nd
2 ? 0/1 0/1 Radial crack
rd
3 ? 47/ 49/ Radial crack
4th 28 27/28; 29/30 28/29; 30/31 Radial crack
AE2
DL + LL DL
AE 1
AE3
22kN 28kN
AE5
AE6
AE4
(?) (?)
Figure 1.9: Arch G: Failure mechanism
Figure 1.10: Arch G: Failure mechanism

Deflections, measured at and span show slight change in curvature around

15kN. The first significant change in deflection is shown around 27kN, shortly before
opening of the 4th hinge and four-hinge mechanism failure. It is possible for the 1st
and 2nd hinges to have opened around 15kN although they only became visible under
higher loads. Deflection measurements therefore do not give clear indication of the
presence of cracks or hinges and can not be directly related to the condition of the
structure.
Figure 1.11: Arch G: Load-Deflection
Surface strain was measured in 5 sections along the arch and 4 locations in each
section as shown in Figure 1.12. Strain measurements against the load at the
span and span is shown in Figure 1.13. Clear change in strain at the span by
gauges B, C and D form 14kN load indicates a radial hinge opening from the intrados
way through the arch section. No change was recorded by strain gauge A at the
top of the section. Changes in strain and deflection (Figure 1.11) around 14kN load
indicate the opening of the first hinge, however the first hinge was visually observed
only around 22kN.
The final change in strain is shown at 27kN which again agrees well with the changes
in deflection. The fourth hinge is therefore likely to have opened at 27kN which
caused the arch to become unstable at 287kN.
Strain measurements have therefore provided useful information regarding changes
in behaviour and opening of hinges.
A E
B F
C G
D H
Figure 1.12: Arch G: Load-Strain

200
0
Microstrain
-200
-400 A
B
-600 C
D
-800
0
10
12
14
16
18
20
22
24
26
28
Load (kN)
200
E
150 F
G
100 H
Microstrain
50
-50
-100
0
10
12
14
16
18
20
22
24
26
28
Load (kN)
Figure 1.13: Arch G: Load-Strain
During load application 6 acoustic emission sensors were additionally attached to the
arch surface. For AE sensor location see Figure 1.9. Recorded emission is shown in
Figure 1.14 for amplitude vs. load. Unfortunately the amplitude limits were set
reasonable high (55dB), hence no information was recorded for low-level emission.
The first significant emission was recorded by AE1 at the span at 14kN. At 22-
23kN AE3 indicated large emission and associated radial crack opening around the
span which only became noticeable with the naked eye around 28kN when the
arch failed.
Amplitude (dB) Amplitude (dB) Amplitude (dB)
100 100
AE1 AE4
80 80
60 60
40 40
20 20
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
100 100
AE2 AE5
80 80
60 60
40 40
20 20
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
100 100
AE3 AE6
80 80
60 60
40 40
20 20
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Load (kN) Load (kN)

Figure 1.14: Arch G: AE Amplitude vs. Load
1.4.1.3 Arch M (5m span)

Arch M was the first 5m span arch to be tested under static loading. A radial hinge
opened during dead load application at the span (at 20kN dead load) through the
bottom two rings but did not propagate any further during load application. No further
damage was observed the arch failed by sudden ring separation under 30kN live
load between the 1/8 and span (see Table 1-6,
Figure 1.15 and Figure 1.17). It is interesting to note that ring separation did not
originate from the initial radial crack but developed separately a short distance away
from it.
After failure the live load was removed and the arch was re-tested to identify the
archs residual strength. The maximum load that could be applied to the damaged
arch was 12.5kN , 42% of the original capacity. The arch collapsed as a four-hinge
mechanism under re-applied loads (see Figure 1.18).
Table 1-6. Arch M: Crack history
Type
(kN) Intrados Middle Extrados
0 24/25 25/26 - Radial crack
10-21; 12-22; -
30 Ring separation
- 12-22 13-23
DL + LL DL
AE3
AE2
AE4
AE1
30kN
DL
AE6
AE7
AE5
AE8
Figure 1.15: Arch M: Crack locations and AE positions
Figure 1.16: Arch M: Failure mechanism

Figure 1.17: Arch M: Ring separation failure
Figure 1.18: Arch M: Collapse mechanism
8 acoustic emission sensors were attached to the arch surface during load
application (see
Figure 1.15 for gauge locations AE1-AE8 and Figure 1.19 for AE amplitude vs. load
recordings). Although no visual signs of damage could be seen until the arch failed
by sudden ring separation under 30kN, AE recordings showed intense activity by
sensors AE1 and AE5 from the start of live load application. Location of intense AE
activity coincided exactly with the location of final ring separation and identified the
Amplitude (dB) Amplitude (dB) Amplitude (dB) Amplitude (dB)
potential areas of damage.
Load (kN) Load (kN)
Figure 1.19: Arch M: AE amplitude vs. Load

1.4.2 Weak brick arches
1.4.2.1 Arch T (5m span)

Arch T was built from weak bricks to represent arches with deteriorated masonry.
During static loading a radial hinge opened under 14kN at the span through the
bottom two rings (see Table 1-7, Figure 1.20 and Figure 1.21). The arch failed by ring
separation between the and span under 26kN live load (87% of the capacity of
the 5m strong brick arch). After ring separation failure the load was further applied up
to 31kN during which further radial and diagonal cracks opened as shown in Figure
1.20. It is interesting to note that both strong and weak 5m arches developed a radial
hinge at the span and failed by ring separation.
Table 1-7. Arch T: Crack history
Type
(kN) Intrados Middle Extrados
14 22/23 23/24 - Radial crack
26 27 - 38 28 - 39 - Ring separation
28 17 - 21 16 - 20 - Ring separation
2/3; 2/3;
30 - Radial crack
38/39 40/41
31 - 24/25 25/26 Ring separation
DL + LL DL
31kN 30kN
28kN
26kN
AE4
AE5
AE3
AE6
AE2
AE7
30kN 14kN
AE8
Figure 1.20: Arch T: Failure mechanism
Figure 1.21: Arch T: Failure mechanism

7 acoustic emission gauges were attached to the arch intrados (for AE gauge
locations see Figure 1.20). AE amplitude vs. load recordings showed increased
activity form the start of load application by sensors AE2 and AE3 which clearly
identified the radial crack opening at 14kN in their vicinity. Sensors AE2, AE3 and
AE4 also show increased activity from ca. 22kN, well in advance of ring separation
failure at 26kN load.
Figure 22: Arch T Amplitude vs. load

1.5 Cyclic loading tests
1.5.1 Strong brick arches
1.5.1.1 Arch C (3m span)

Cyclic loading was applied to arches alternatively over the and spans at 2Hz
frequency to represent the flow of traffic over the bridge. Arch C was loaded cyclically
at the and spans by 14kN live load (approx. 50% of the static capacity of 3m
span strong brick arches). During initial load application radial hinges opened under
9kN at the span and left hand abutment and under 11kN at the right hand
abutment (see Table 1-8 and Figure 1.23). 14kN cyclic load was applied for 23000
cycles, when the arch failed by sudden ring separation in the middle section of the
arch (see Figure 1.24).
Cyclic loading was subsequently re-applied to gain information on the reserve
capacity of the damaged arch. Sliding (shear) failure developed rapidly at the
span at a pre-existing radial hinge (Figure 1.25) as the loading jack repeatedly forced
the arch section downwards. The arch collapsed after 150 cycles by the formation of
hinges and sliding at the span (see Figure 1.26). Capacity of the damaged arch
was determined by the residual frictional strength of the brick - mortar interface.
Table 1-8. Arch C: Crack history
Type
12/13 - Radial crack
9
0/1 0/1 Radial crack
11 48/ 50/ Radial crack
19/28 20/29 Ring separation
14 (23000 cyc.)
47/ 50/ Radial crack
14kN
14kN (23000 c.)
(23000 c.)
9kN 14kN
9kN 11kN
(23000 c.)
(c. = cycles)
Figure 1.23: Arch C Failure mechanism

Figure 1.24: Arch C: Ring separation
Sliding
Figure 1.25: Arch C: Sliding failure
Figure 1.26: Arch C: Collapse mechanism

1.5.1.2 Arch E (3m span)

During the removal of centering after constructions a number of radial cracks were
noticed on the arch (shown as 0kN in Table 1-9, Figure 1.28 and Figure 1.29). Cyclic
loading was applied at 12kN (43% of the static capacity of 3m span strong brick
arches). No sign of damage was noted until the arch failed by sudden ring separation
after 23,000 load cycles between the two abutments. The arch collapsed 2,320
cycles after ring separation by formation of five hinges (see Figure 1.30).
Table 1-9. Arch E: Crack history
Type
- 0/1
- 50/
14/15; 16/17 -;- Radial crack
DL (0kN) 24/25 25/26
33/34 35/36
23 - 25 24 26 Longitudinal crack
DL (8kN) - 24/25 Radial crack
12 (23,000 cycles) 16/32 17/34 Ring separation
12kN
(23000 c.)
DL
DL DL
Figure 1.27: Arch E: Failure mechanism
Figure 1.28: Arch E: Failure mechanism

8kN DL
Drying cracks
(0kN)
12kN (23,000 cyc.)
Figure 1.29: Arch E: Crack locations
Figure 1.30: Arch E: Collapse mechanism

1.5.1.3 Arch O (5m span)

During dead load application radial cracks opened at the and span (see Table
1-9). Cyclic loading was applied for 106 cycles at a series of load levels starting from
a relatively small (6kN) load. No damage was observed under 6, 8, 10, 12, 13kN
loading series. Under 14kN the existing radial cracks at the and spans extended
slightly however no further damage occurred for the rest of 106 cycles. The arch
finally failed under the 18kN loading series with ring separation in the middle section
of the arch after 340,000 cycles (see Table 1-9, Figure 1.31, Figure 1.32 and Figure
1.33).
12kN cyclic load was subsequently re-applied until collapse occurred. Unlike the well
known four-hinge mechanism, the arch collapsed by forming five hinges as shown in
Figure 1.34). The fifth hinge was a function of the loading system which vertically
confined the arch and did not allow sufficient upwards movement when loading was
alternatively released at the and spans.
Table 1-10. Arch O: Crack history
Live load (kN) [% static Crack location
Type
capacity*] Intrados Middle Extrados
22/23 23/24 - Radial crack
0
55/56 - - Radial crack
6 [20%] - - - -
8 [27%] - - - -
10 [33%] - - - -
12 [40%] - - - -
13 [43%] - - - -
24-25; 25-26;
14 [47%] 500,000 cyc. - 36-43; 57-44; Longitudinal crack
56-57 57-59
16 [53%] - - - -
18 [60%] 340,000 cyc. - 36-43; 37-44 Ring separation
* % of the 30kN static capacity of the 5m strong brick arch M (section 1.4.1.3)
18kN
(40,000c.)
0kN
? ?
Figure 1.31: Arch O: Failure mechanism (18kN)

Figure 1.32: Arch O: Failure mechanism (18kN)
Figure 1.33: Arch O: Ring separation (18kN)
Figure 1.34: Arch O: Collapse mechanism
8 acoustic emission sensors were attached to the arch intrados during the loading
series (for AE sensor locations see Figure 1.31). The average released absolute
energy per cycle is summarised for the beginning and end of each 106 cycles for
each load level (i.e. for 0 cycle and 106 cycles) in Figure 1.35. The released energy
level was relatively low for 10kN and 12kN and started to increase from the start of
14kN load application. Under 18kN the released energy increased further until failure
occurred. With the help of acoustic emission recordings the 14kN load level could be
identified as the fatigue (permissible) limit for the arch, above which accelerated
damage propagation took place and caused eventual failure.
Abs. Energy (aJ) / Cycle
120
100
Permissible
separation
80
Ring
limit
60
40
20
0
10kN
10kN
12kN
12kN
14kN
14kN
16kN
16kN
18kN
18kN
Loading history
Figure 1.35: Arch O: AE Absolute energy per Cycle vs. Loading history
1.5.2 Weak brick arches
1.5.2.1 Arch S (5m span)

Similarly to the 5m weak brick arch O (described in 1.5.1.3) radial cracks (as well as
a short section of ring separation) opened at the and span in arch S during live
load application (see Table 1-11 and Figure 1.36). Cyclic loading was applied at
12kN (46% of the 5m weak brick arch described in section 1.4.2.1). Sudden ring
separation occurred between the and span after 2400 cycles. During further
cyclic loading ring separation extended from the to the span after 20,000 cycles
(see Figure 1.36 and Figure 1.37).
Table 1-11. Arch S: Crack history
Type
capacity**] Intrados Middle Extrados
0 55/56 57/58 - Radial crack
- 55-57 57-59 Longitudinal crack
2,400 cyc. - 40-55 42-57 Ring separation
- 25-40; 25-42; ---
Ring separation
26-39 27-41 -
12 [46%]
20,000 cyc. 51/52; 55/56;
57/58;
53/54; 56/57; Radial cracks
58/59
55/56 57/58
** % of the 26kN static capacity of the 5m weak brick arch T (section 1.4.2.1)
12kN 12kN
(20,000c.) (2,400c.)
0kN
Figure 1.36: Arch S: Failure mechanism


1.5.2.2 Arch U (5m span)

Similarly to the previous 5m span arches, initial load application was accompanied by
radial cracks at the an span (see Table 1-12 and Figure 1.39). A series of load
levels were applied for 4000 cycles starting from 10.5kN (40% of the capacity of the
5m weak brick arch). Almost no damage occurred under 10.5, 13.5, 14.5, 15.5 and
16.5kN load series. Ring separation occurred under 16.5kN between the span and
nearest abutment after 680 cycles (see Figure 1.41) which was determined failure of
the arch. After ring separation loading was maintained for an other 150,000 cycles
without any further damage propagation. During further load increase the arch finally
failed under 25.5kN (98%) by full ring separation under 25.5kN (see Table 1-12).
Table 1-12. Arch U: Crack history

Type
capacity**] Intrados Middle Extrados
0
53/54 - - Radial crack
10.5 [40%] - - - -
13.5 [52%] - 56/57 - Radial crack
14.5 [56%] - - - -
15.5 [60%] - - - -
16.5 [64%] (680 cyc.) 58-75 60-78 - Ring separation
17.5 [67%] - - - -
18.5 [71%] - - - -
19.5 [75%] - - - -
20.5 [79%] - - - -
21.5 [83%] - - - -
22.5 [87%] - - - -
23.5 [90%] - - - -
24.5 [94%] - - - -
- 60-71 63-75 Ring separation
25.5 [98%]
- 40/41 42/43 Radial crack
** % of the 26kN static capacity of the 5m weak brick arch T (section 1.4.2.1)
25.5kN 13.5kN
25.5kN
0kN 16.5kN
(680c.)
Figure 1.39: Arch U: Failure mechanism
Figure 1.40: Arch U: Failure mechanism
Figure 1.41: Arch U: Ring separation

1.6 Summary
1.6.1 Static loading summary

Under static loading 3m and 5m span arches failed by different mechanisms: 3m
span (2-ring) arches failed by four-hinge mechanism and the 5m span (3-ring) arches
failed by ring separation (see Figure 1.42). During the development of four-hinge
mechanism hinges opened gradually, starting from a low load level, although they
often became visible under significantly higher loads. Deflection became non-linear
only after the opening of the 3rd or 4th hinge (see Figure 1.43a and b). Failure was
predictable through visual signs of hinge opening and non-linear deflection response.
In 5m span arches ring separation occurred suddenly, without any visual warning or
changes in the deflection response (see Figure 1.43c and d). In both 5m span arches
a radial hinge opened at a low load level under at the span directly under the live
load. Vertical deflection under the span was significantly greater for the weak brick
arch which is likely to be due to the significantly lower Youngs modulus for weak
brickwork (Youngs modulus of weak brickwork was ca. 1/3 of strong brickwork).
DL + LL DL DL + LL DL
23kN 22kN 28kN

6.5kN
14kN (?) (?) (?)

a) Arch A (Strong) b) Arch G (Strong)
DL + LL DL DL + LL DL
30kN
DL
14kN
c) Arch M (Strong) d) Arch T (Weak)

Figure 1.42: Static failure modes
a) Arch A (Strong) b) Arch G (Strong)
c) Arch M (Strong) d) Arch T (Weak)

Figure 1.43: Load vs. Deflection
Despite the different failure mechanisms, average static capacity was very similar for
3m and 5m span arches: 28.5kN and 28kN for the 3m and 5m arches respectively.
Capacity of the 3m arches was determined by the flexural tensile strength of masonry
that caused formation of a four-hinge mechanism, while capacity of the 5m arches
was determined by the shear strength of masonry and caused ring separation failure.
Although deflection can easily be measured on arch bridges, measurements have
only shown changes in the archs behaviour well after the development of significant
cracks. Deflection measurements may therefore not be reliable indications of the
archs condition. Strain measurements were able to identify crack development
immediately after their occurrence but only when measured in the appropriate arch
sections. Acoustic emission (AE) technique was able to indicate internal damage
propagation prior to crack opening and accurately identify future areas of damage. It
showed great potentials for condition assessment of masonry arch bridges.
1.6.2 Cyclic loading summary

Cyclic load caused all 3m and 5m span arches to fail by sudden ring separation (see
Figure 1.44). During initial load application strong arches generally developed four
radial hinges (at the 0, , and 1 span, see Figure 1.44a, b and c), while weak
brickwork arches only developed two radial hinges (at the and span as shown in
Figure 1.44d and e). The flexibility of weak brickwork provided sufficient movement to
replace two releases (hinges) compared to strong brickwork arches.
Failure under cyclic loading occurred suddenly, without any warning visually, by
deflection gauges or by strain measurements. Acoustic emission monitoring was
however able to indicate increased emission level well in advance of failure (see
Figure 1.35), warn of increased internal damage development immediately prior to
failure and could help identify the archs permissible limit state.
14kN (23000 c.) 12kN (23000 c.)
Initial loading Initial loading
a) Arch C (Strong) b) Arch E (Strong)

18kN (40,000c.)
Initial loading
c) Arch O (Strong)
12kN (20,000c.) 12kN (2,400c.) 16.5kN
(680c.)
Initial loading Initial loading
d) Arch S (Weak) e) Arch U (Weak)

Figure 1.44: Cyclic failure modes
1.6.3 SN curves
The summary of the arch test series is shown in Table 1-13. Arches generally
developed a number of radial cracks relatively early during load application that could
be used for identifying likely modes of failure. Figure 1.45 shows a diagrammic
summary of visual defects with their likely modes of failure from the previously
described tests.
Table 1-13. Arch test summary
% of average
Loading
Max. load No. Failure

Arch Brick Span static
(kN) cycles mechanism
capacity
A 29 100% 1 Four-hinge
3m
G Strong 28 (28.5kN) 1 Four-hinge
Static
M 30 1 Ring separation
5m 100% (28kN)
T Weak 26 1 Ring separation
C 14 50 23,000 Ring separation
3m
E Strong 12 43 23,000 Ring separation
Cyclic
O 18 60 40,000 Ring separation

S 5m 12 46 2400 Ring separation
Weak
U 16.5 64 680 Ring separation
Defects Location Likely failure mode
1 radial crack Ring separation
2 radial cracks ,
Ring separation
3 radial cracks 0, , 1
4-hinge mechanism
0, , 1
Abutment movement
4 radial cracks 0, , , 1
Ring separation
0, , , 1
4-hinge mechanism
0, , , 1
Sliding
Figure 1.45: Arch defects and likely failure modes

Under static loading both 3m and 5m span arches failed at similar load levels but by
different mechanisms. Under long-term fatigue loading the permissible limit for the
arch series was around 40% of the static capacity, above which the arches failed
within a relatively low number of cycles.
A possible way of relating load levels to the number of cycles an arch can carry
before failure is with the help of SN curves. Arches can fail by a number of
mechanisms, e.g. four hinge mechanism, ring separation, sliding, etc., all of which
should be represented by individual SN curves. Figure 1.46 shows an indication of
SN curves for ring separation failure from the previously described arch tests series.
Stresses are expressed as percentage of the maximum static stress against and are
related to the number of cycles at failure. The two different brick qualities (Class A
engineering and Britley Olde English bricks) used in the test series are graphically
identified and bands for different brickwork qualities (A, B, C) are suggested. Large
volumes of further test data are now needed to develop SN curves for various failure
modes and masonry qualities.
Subsequently the SN curves can be incorporated into the newly developed SMART
assessment technique to provide a holistic assessment tool that can identify the
bridges service life under any working stress level and their likely modes of failure.
The SN curves can also be used to indicate the possible sensitivity of the bridges life
expectancy to changes in working stresses levels. If for example the shear stresses
increase from 50 to 55% of the max. static stress, life expectancy of the bridge may
reduce almost to its quarter (from 300,000 to 80,000 cycles) as shown in Figure 1.46.
SN curves can also indicate the permissible limits for each failure mode and material
quality below which no residual damage is likely to occur to the arch and loading can
be applied for a theoretically unlimited number of cycles without requiring
reinforcement or strengthening works.
100
Class A
Engineering bricks
Max. static stress (%)
80
Britley Olde
English bricks
60
55
50 Permissible limit
40
A
20 B
0
80,000
300,000
1,E+00
1,E+02
1,E+04
1,E+06
1,E+08
1,E+10
1,E+12
No. cycles
Figure 1.46: S-N curves for ring separation failure

2. Soil-structure interaction in masonry arch bridges

2.1 Background
Although it is well accepted that soil-arch interaction has a significant influence on the
load carrying capacity of many masonry arch bridges, the complex nature of the soil-
arch interaction is not well understood. In the past decades comprehensive research
has been carried out on masonry arch bridges. However, the performance of the
backfill has generally not been the focal point and almost all laboratory bridges tested
to date have been backfilled with granular material bridges (e.g. Page 1993, Fairfield
and Ponniah 1993, Melbourne and Gilbert 1995). Additionally, rigid abutments have
been adopted in most laboratory tests which are less representative to those found in
practice.
Recent intrusive investigations performed on local authority owned bridges in the UK
have frequently identified that abutments are relatively insubstantial (e.g. of
comparable thickness to the arch barrel), and that clay backfill is present. At present
no experimental studies of the performance of clay backfilled arches seems to exist.
Additionally it has frequently been found that abutments are of comparable thickness
to the arch barrel (i.e. are relatively insubstantial). Unfortunately conventional arch
assessment methods have not been calibrated using clay backfill materials, or using
abutments which are not rigid. This makes realistic assessment difficult at the present
time.
The principal aim of the investigations described is to gain an improved
understanding of soil-arch interaction. This is being achieved by conducting a series
of carefully controlled laboratory soil-arch tests to obtain a high quality data-set,
which can be used to calibrate numerical models.
2.2 Objectives
Three bridges have been tested to failure to date, with the following objectives:
To test out the performance of the apparatus (tank, imaging and
instrumentation).
To determine the influence of abutment fixity on arch behaviour.
To provide a benchmark test result.
To better understand the behaviour of masonry arch bridges with clay backfill.
The first test bridge was designed to be similar to the 3m span bridges tested at
Bolton in the 1990s (Melbourne and Gilbert 1995), thereby permitting direct
comparison. However, unlike the Bolton bridges, which had been constructed
between rigid abutments potentially movable abutments were specified and
furthermore the walls of the plane-strain testing tank marked the edges of the bridge,
rather than the brickwork spandrel walls used previously.
The second test bridge was designed to be identical to the first with the exception
that fill material below the level of the crown of the arch was replaced with soft clay,
representative of that found in some local authority owned bridges in the UK.
Crushed limestone was used above the crown to reflect the fact that competent near
surface road/sub-base material is normally present in real bridges. Additionally, had
this not been used then a local failure of the soil in the vicinity of the applied load
would have been likely.
The third bridge was successfully tested on 14 June 2007 and the test results are
currently being processed and should be completed before the end of the contract
(results, where available, are included in this document). The third test bridge was
filled with the same limestone used for the first bridge test. Unlike Bridge 1, twelve
pressure cells were incorporated in this bridge, enabling valuable additional
information to be obtained.
2.3 Experimental programme
2.3.1 Test rig

A purpose-built tank (8.3 m long 2.1 m high) to house an arch and surrounding fill
material was designed with the original requirements being as follows:
Provision of predominantly transparent sides to one face, to enable images of
the soil failure mode to be captured (images from digital camera to be
subsequently fed into image processing software, converted into displacement
vector plots, and then compared with output from numerical models).
Tank to be large enough to house a 3m span segmental arch (span:rise of 4:1)
and a semicircular arch, with the soil failure mechanism not unduly affected by
tank boundaries.
Provision of high stiffness and frictionless side-walls so as to properly model
plane-strain conditions.
Provision of abutment blocks which could be fixed or left free to slide.
The test tank with surrounded scaffolding is shown in Figure 2.1. The dimensions of
the test rig are shown in Figure 2.2. The frame was primarily constructed from heavy
duty steel I sections (406 x 140 x 39UB, Grade S275) to ensure adequate stiffness so
that the plane strain conditions were maintained under load. The frame was designed
to ensure no end effects affected the results, taking into account the anticipated
failure mechanisms. The length of the rig necessitated inclusion of several tie bars
across the top and bottom of the frame to provide adequate lateral stiffness. The
frame supported stiff walls consisting of 50mm thick plywood on the ends and along
one side. On the other side, 50mm thick acrylic windows were incorporated in order
that soil kinematics could be observed. Both walls had a further 6mm layer of acrylic
sheet placed on their internal faces.
Figure 2.1: Test rig with scaffolding
Plan view
Elevation
Figure 2.2: General arrangement of test rig (dimensions in mm)
2.3.2 Bridge geometry and material properties
2.3.2.1 Bridge geometry

The bridge tests were carried out in the purpose-built tank described in section 2.3.1.
The arch barrels were segmental and had a span of 3m, with a nominal span to rise
ratio of 4:1 and over an average width of 1010 mm. Each arch barrel consisted of two
rings and alternate courses comprised headers. These were used to prevent ring-
separation occurring during the test.
The backfills employed were compacted to a depth of 305mm above the arch crown.
The average width was 1045 mm.
Figure 2.3 to Figure 2.5 show the bridge dimensions which have been tested to date.
215
305
1665
750
395
550 180 3373 550
730 322 3000 325 3923
8300
elevation
Figure 2.3: Bridge1 dimensions (mm)
215
305
1665
750
395
1930 322 3000 325 2723
8300
elevation

305
1665
750
395
3318 325 3000 322 1335
8300
2.3.2.2 Material properties

Class A Engineering bricks described as Nori bricks were used in the construction
of the arch barrel. The mean properties of the bricks are given in Table 2-1.
The mortar 1:2:9 (cement:lime:sand) mix by volume was used throughout the arch
barrel. The cubes were cured under the same conditions as the arch barrel. The
mean properties as determined from five 100mm cubes are presented in Table 2-1.
Five brickwork prisms were constructed using the same materials and workmanship
as those used in the construction of the arch barrel. The prisms were built five
courses high. The average density and compressive strength are given in Table 2-1.
The first and third bridges (Bridge1 and Bridge3) were backfilled with MOT Type 1
graded crushed limestone sourced from Tarmac Central Ltd - Holme Hall Quarry. A
total of 10 shear box tests were carried out on the material for Bridge1. The average
density and strength are presented in Table 2-1. The shear properties did display
stress dependence. Dilation is clearly suppressed at higher stresses.
The second bridge (Bridge2) was backfilled with clay to crown level with 305mm of
compacted crushed limestone above. The clay was supplied by Marchington Stone
Ltd and was described as a firm red-dish brown slightly sandy CLAY with occasional
gravel. The suppliers determined index properties were as follows: natural moisture
content 15%; optimum moisture content 9%; Liguid Limit 29%; Plastic Limit 12%. The
average moisture content on material placement was 13.3%. The average density
and shear strength from undrained tests on eight samples taken after the test are
presented in Table 2-1.
Table 2-1. Material properties
Material Type
Masonry units Compressive strength (N/mm2) 154
Unit weight (kN/ m3) 23.2
Mortar Compressive strength (N/mm2) 1.9
Unit weight (kN/ m3) 14.4 15.4
Masonry Compressive strength (N/mm2) 24.5
Unit weight (kN/ m3) 21.6
Backfill Unit weight (kN/ m3) Limestone: 19.1
Clay: 22.1
(degrees) Limestone: 46.4
2
Cohesion (kN/mm ) Limestone: 22.4
Clay: 78
2.3.3 Bridge construction
2.3.3.1 Abutments
Two reinforced concrete abutments on which the arch barrel were built were fixed
3000 mm apart, parallel to each other. The dimensions of the two abutments are
shown in Figure 2.6.
Each abutment comprised two parts; a lower section that was bolted to the structural
strong floor, and an upper section that could slide should the forces be large enough
to overcome the bond/friction between the blocks.
230 154
215
13
530
530
20
395
322 325
south abutment details north abutment details

(width 999mm) (width 999mm)
Figure 2.6: Construction details of the abutments (dimensions in mm)
2.3.3.2 Centering
The arch was constructed on custom made steel centering (as shown in Figure 2.7)
on which 101.6mm x 50.8mm x 1000mm planks were placed, in turn covered by a
sheet of plastic in order to minimise bonding of the masonry to the planks and
facilitate easy removal of the centring. Each curved steel beam was supported by two
individual stacks of bricks.
215
crown
750
3000
Figure 2.7: General arrangement of centering and arch barrel
2.3.3.3 Arch barrel

Each of the arch barrel tested was segmental of 3m span, with a nominal span to rise
ratio of 4:1. It consisted of two rings and was constructed over an average width of
1010mm. Figure 2.7 shows the arch barrel under the construction and the
arrangement of the arch barrels. Headers were used to prevent ring-separation
occurring.
2.3.3.4 Tank
Following construction of the arch, the test tank was assembled around the arch. The
average width of the test rig was 1045mm while the average width of the arch was
1010mm. Tie bars were placed across the top and bottom of the frame to provide
requisite lateral stiffness.
In order to prevent fill falling between the gap between the arch and the test rig walls,
strips of closed cell foam were hot glued along the edges of the arch extrados so as
to span the gap (see Figure 2.8). The flexible foam would accommodate any minor
lateral arch movements while retaining the fill above.
Figure 2.8: Gap covered with foam and attached latex marked
2.3.3.5 Wall friction reduction

The tank was design to be sufficiently stiff to provide plane strain conditions. This
meant that significant confining pressures might develop between the backfill and the
tank sides. The consequence of these pressures would be to develop significant
frictional forces.
To minimise side wall friction, the full faces of the 6mm perspex sheets were covered
in a layer of silicone grease followed by a 0.33mm thick latex sheet. Fang et al.
(2004) discuss a range of friction reducing measures in detail. Of those considered,
greased latex offered the lowest friction angles for normal stresses greater than 5
kPa. It was considered that such stresses would dominate in such a large model
employing compacted backfill and with high stresses beneath the loading platen. For
normal stresses greater than 10kPa, Fang et al. indicate interface friction angles of
less than 2. Figure 2.8 shows the silicone greased latex attached to the plywood and
acrylic walls and secured with tape at top.
2.3.3.6 Backfill
The fills were compacted in layers using vibrating compaction plates. Sensitive areas
adjacent to the walls and the arches were compacted with a hand rammer.
For the fill comprised MOT Type 1 graded crushed limestone (Bridge1 and Bridge3)
the required weight for each layer was loaded into a crane mounted hopper for
transfer to the rig and spread evenly using a shovel to the required thickness, and
then compacted to the required specification using 10kN (1t) whacker plate.
The use of greased latex sheeting unfortunately reduced the visibility of the fill
material slightly. In order to ensure the images captured had sufficient contrast to be
processed using the GeoPIV software, 50mm gaps on the inside of both walls were
set up. The gaps between the plates and the walls were filled with coloured fill, as
shown in Figure 2.9, while normal fill was placed in the large central gap between the
plates. Figure 2.10 shows the completion of the filling of limestone for Bridge1 and
Bridge3.
Figure 2.9: Gaps set up and filled with coloured fill to enhance the post test image
analysis
Bridge1 Bridge3
Figure 2.10: Filling completed for limestone filled bridges
Filling and compacting of clay on such a large scale is challenging. Trial compactions
were carried out involving hand compaction in a compaction mould (Proctor
Compaction Test) and compaction on small scale 1m2 box using a vibrotamper at the
north end of the fully assembled tank. Guided by the experience obtained from these
tests, a filling methodology was established. The clay was first wetted up in batches
as appropriate to the required consistency and thoroughly mixed using the back-
acting arm of an excavator. It was then transferred to the test rig using the excavator
bucket and spread evenly using a shovel to the required thickness. Each layer was
then compacted using a 10kN vibrating compaction plate suspended from a crane to
ease handling. Sensitive areas adjacent to the walls and arch itself were compacted
with a hand rammer. Each pressure cell was hand covered using clay. No
compaction was employed directly adjacent or over the cells.
During clay filling, small soil samples were taken at regular intervals for moisture
content testing. A total of 95 samples were collected and the average moisture
content was 13.4%. Readings from a pocket penetrometer were also taken at regular
distance after compaction of each layer during the whole filling process.
To protect the clay from drying out after compaction into the tank, several cloth and
plastic sheets were used to cover the clay during and after filling. Figure 2.11 shows
the procedures of clay filling and Figure 2.12 shows the completion of the filling of
clay.
Figure 2.11: Bridge2 clay filling procedures
Figure 2.12: Bridge2 clay filling completed and covered
The Bridge2 filling was finished with two layers of limestone on the top of the clay.
2.3.4 Bridge instrumentation
2.3.4.1 Deflection
Linear Variable Differential Transformer (LVDT) type displacement transducers were
placed beneath the intrados of the arch barrel and the top parts of the abutments to
measure its movement. The accuracy of these gauges was about 0.01mm.
Additionally, LVDT and/or mechanical dial gauges were being used to monitor lateral
movements of the test rig, to ensure that the conditions of plane strain are met.
Figure 2.13 and Figure 2.14 show the positions of these gauges for Bridge1 and
Bridge3 respectively.
Figure 2.13: Layout of displacement transducers monitoring the arch and the test rig
(Bridge1)
G-3/4L G-1/4L
West East
abutment G-West G-East abutment
LV9 - Top LV12 - Top
LV10 - Middle LV13 - Middle
LV11 - Bottom LV14 - Bottom
West abutment
219
East abutment
750
920
LV7 - Top LV15 - Top

LV8 - Bottom Dial gauge - Bottom
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
LV4 - Top LV1 - Top

LV5 - Middle LV2 - Middle
LV6 - Bottom LV3 - Bottom
Figure 2.14: Layout of displacement transducers monitoring the arch and the test rig
(Bridge3)
2.3.4.2 Earth pressure

Wall-Surface Soil Pressure Transducer BER-A-12S (200kPa and 500kPa) were used
to monitor earth pressure. A total of four pressure cells were embedded into the arch
extrados in Bridge2 and twelve in Bridge3. The positions of these pressure cells on
Bridge3 tests are shown in Figure 2.15
PC12 PC11 PC10 PC9 PC8 PC7 PC6 PC5 PC4 PC3 PC2 PC1
West abutment
East abutment
Figure 2.15: Positions of pressure cells on arch extrados (Bridge3)
2.3.4.3 Imaging
The testing tank provides 14 bays between the steel columns along its length. Of
these, the middle 12 bays incorporate acrylic windows. In order to capture soil
displacements, a set of six SONY DSC-V1, 5 MegaPixel, digital cameras are set up
such that each camera images a pair of bays (as shown in Figure 2.16). During
testing, and following the application of each load increment, the cameras can be
remotely triggered in quick succession to capture the images. Illumination from above
each bay with a halogen lamp has proved essential to ensure good image quality and
to minimise reflections.
Figure 2.16: Setting up cameras for soil imaging
2.3.4.4 Acoustic monitoring

A total of 8 acoustic gauges were placed beneath the intrados of the arch barrel, as
shown in Figure 2.17, to enable real-time fracture activity to be recorded.
South North
abutment abutment
Figure 2.17: Layout of Acoustic gauges (Bridge1)
2.3.5 Test Procedure and loading arrangement

The instrumentation, as described in Section 2.3.4, was monitored throughout all the
tests.
The load was applied in increments of about 5 10 kN when the increment of radial
displacement at quarter span was less than 1mm. If the increment was over 1mm,
load was control by displacement increment. For each increment, the images were
taken, the deflection of the arch at 1/4L was checked, and the test rig (tank)
movement was checked through the dial gauges and LVDT's.
Two hydraulic jacks supported from a steel reaction frame were used to apply a line
load onto the surface of the backfill above the quarter point of the arch barrels. The
load was applied vertically onto the surface of the backfill through a steel loading
beam (base 920x219) resting on a same size wood base which was placed on the
surface of the backfill. The steel reaction frame was bolted to the strong floor. Figure
2.18 shows the loading arrangement used for Bridge1 test.
reaction beam
loading beam
backfill wood base
arch barrel
Section A-A
A
north abutment
south abutment
219
1045
wood base
920
750
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
A
Figure 2.18: Loading arrangement (Bridge1)
The third limestone filled bridge (Bridge3), which was tested on 14 June 2007, was
carried out in the newly relocated laboratory. Figure 2.19 shows the similar loading
arrangement for Bridge3 test.
Figure 2.19: Loading arrangement for third bridge (Bridge3) test
2.3.6 Test results
2.3.6.1 Visual observations

The arch barrels of the tested bridges were inspected prior to commencing the tests.
No cracks were found except a few minor shrinkage cracks in Bridge2. However,
because not all of the joints on the intrados were fully filled, it made inspection for
cracks on the intrados difficult.
The first cracks were always found in the arch intrados immediately under the load
(quarter span).
All the three bridges tested ultimately collapsed in four-hinged mechanisms, as

South North
abutment abutment
Bridge 1 (limestone fill)
South North
abutment abutment
Bridge 2 (clay fill)
West East
abutment abutment
Bridge 3 (limestone fill)
Figure 2.20: Collapse mechanism of arch barrels
The limestone filled (Bridge1) failed at a load of approximately 125kN, the clay filled
bridge (Bridge2) failed at a load of approximately 90kN, and recently tested Bridge 3
failed at the load of approximately 148kN.
Figure 2.21 shows the bridge and surrounding limestone backfill close to the point of
collapse from the observation of the large windows.
Figure 2.21: Bridge1 (Limestone-filled bridge) close to the point of collapse
Cracks, especially at/near both abutments, were clearly seen after one side of the
walls was taken away. Figure 2.22 shows the cracks in the arch barrel after Bridge3
test. Figure 2.23 shows the standing bridge after one side of the wall was removed.
Figure 2.22: Cracks in arch barrel ( Bridge3 test)

Figure 2.23: Bridge3 after one side of the wall removed
Whilst at collapse hinged mechanisms formed, some abutment movement was

recorded prior to this stage. In the case of the crushed limestone filled bridge
(Bridge1) spreading of the abutment remote from the applied load peaked at 3 mm,
whilst in the case of the clay filled bridge this was significantly greater, peaking at 8
mm. However, in the case of the clay filled bridge rotation about the base of the lower
abutment block remote from the applied load was identified, despite the fact that this
had been bolted to the structural strong floor. In the recently finished third bridge test,
the spreading of the abutment remote from the applied load was 1.8 mm although
initial intention was to fix the abutment for this test. In the case of each bridge,
movement of the abutment closest to the applied loads was small < 1 mm.
Deformations of the test rig were small and within the design limits set (maximum of
0.3 mm, 0.8 mm and 0.6 mm in Bridge1, Bridge2 and Bridge3 respectively).
2.3.6.2 Load-deflection response

Figure 2.24 shows the quarter span deformations from the three bridges tested to
date. It is evident that the bridge backfilled with crushed limestone had a significantly
higher load carrying capacity than the bridge filled with soft clay.
160
Bridge3 (limestone filled)
140
120
Bridge1 (limestone filled)

100
Applied load (kN)
80
Bridge2 (clay filled)
60
40
20
0
0 5 10 15 20 25 30
Radial displacement (mm)
Figure 2.24: Load displacement relationship at quarter span position
2.3.6.3 Arch and soil kinematics

Example images processed using GeoPIV (White and Take 2002) are shown in
Figure 2.25 and Figure 2.26.
Figure 2.25: Load Images of limestone filled bridge with PIV soil displacement
vectors superimposed
Figure 2.26: Images of clay filled bridge with PIV soil displacement vectors
superimposed
Bridge3 results are being processed and will be added

2.4 Finite element modelling
2.4.1 Finite element modelling approach

While the Finite Element Method is widely and successfully used when analysing
both steel and reinforced concrete structures, its utilization in the analysis of masonry
still provokes some disputes in the scientific community. The main reason is due to
the particular characteristic of masonry, which is an anisotropic composite material of
units and mortar. Several earlier developed FE based computer programs do not
adequately describe the complex mechanical behaviour inherent to masonry
elements. In the last decade significant effort has been made to the numerical
modelling of masonry, but the current knowledge about masonry mechanics is
underdeveloped, in comparison with other fields such as concrete or steel. It is still
very difficult to introduce correct constitutive laws for the basic material and to ensure
the convergence of such highly nonlinear problems. Special care must be taken,
therefore, when using the results from finite element analysis of masonry structures.
2.4.1.1 Masonry modelling approach

The existing models designed for the analysis of masonry have been developed in
the last 20 years. A brief description of the masonry modelling approach is discussed
here. A detailed review of the modelling approach has been reported elsewhere
(Wang, 2004).
Masonry is a composite material of bricks and mortar. In general, the approach
towards its numerical representation can focus on the micro-modelling of the
individual components, or the macro-modelling of masonry as a composite as shown
in Figure 2.27.
Figure 2.27: Modelling of masonry structures
Micro-modelling simulates each constituent of the composite material of masonry

separately with its own specific constitutive law and failure criteria. Depending on the
level of accuracy and the simplicity required, micro-modelling can be further divided
into two levels: detailed micro-modelling and simplified micro-modelling.
In the level of detailed micro-modelling, the units and the mortar in the joints are
represented by continuum elements whereas the unit-mortar interface is represented
by discontinuous elements (interface elements) and placed at both areas of adhesion
on either side of the joint, as illustrated in Figure 2.28 (a). In this approach, Youngs
modulus, Poissons ratio and, optionally, inelastic properties of both brick and joint
material are taken into account. In principle, these detailed analyses would allow
consideration of all possible failure mechanism. The disadvantages of this level of
modelling are: firstly, that it requires a highly refined mesh, which has huge
computational cost; and secondly, correct material properties and constitutive laws
for the units and mortar joints are difficult to define.
In the level of the simplified micro-modelling, masonry is represented by continuum
elements where the behaviour of the mortar and unit-mortar interfaces is combined to
give a discontinuous element, as shown in Figure 2.28 (b); with this approach, it is
possible to consider the masonry as a set of elastic blocks bonded together by
potential fracture lines at the joints. In this level of modelling, only one interface
element is used for the mortar joint, including both areas of adhesion, so that
considerably fewer elements are necessary compared with detailed micro-modelling.
A disadvantage here is that Poissons effect of mortar is ignored which will be less
accurate for masonry in compression.
(a) detailed micro-modelling (b) simplified micro-modelling
Figure 2.28: Modelling strategies for masonry (Lourenco 1996)
Macro-modelling considers the composite material of masonry as a smeared

continuum with average stress-strain relationship and a series of global failure
criteria. It does not make a distinction between individual units and joints, but treats
the masonry as an aniostropic composite. The analysis of masonry structures built
from a large number of units and joints can only be carried out with macro-models.
Unlike masonry material modelling where great interests are in the level of micro-
modelling, the modelling methods of masonry arch bridges have mainly concentrated
on global aspects rather than on the simulation of its constitutive materials (Boothby
et al 2001, Fanning et al 2001, Garrity et al 2001, Tao 2002, etc.).
Early analytical studies concentrated on elastic isotropic behaviour, and later macro
models used an isotropic homogenised failure surface similar to those developed for
the analysis of concrete structures. Some of the most popular failure surfaces are
The current study model the brickwork arches based on macro-modelling. The
smeared cracking method is adopted in which the cracking is modelled through an
adjustment of material properties which effectively treats the cracking as a smeared
band of cracks, rather than discrete cracks. The complex behaviour of masonry is
assumed to be isotropic before cracking and orthotropic after cracking. Cracking is
permitted in three orthogonal directions. As concrete element SOLID65 in ANSYS 9.0
was used to model the arch barrel, consequently Willam Warnke failure model
(Willam-Warnke 1975) was adopted for describing the failure surface of masonry
material in a three-dimensional system.
Figure 2.29: Failure models (Chen 1985)
2.4.1.2 Backfill modelling approaches

Different constitutive models have been proposed for soil modelling. These
constitutive models are essentially pressure-dependent plasticity models. The
differences are based on the shape of the yield surface in the meridian plane (either
curved or straight meridians, with or without tension cut-off and/or compression cap),
the shape of yield surfaces in the deviatoric stress plane (either circular or noncircular
yield surfaces), and the use of flow laws (either associated or non-associated flow
rules). The choice of the model to be used depends largely on the kind of the
material, on the experimental data available for calibration of the model parameters,
and on the range of pressure stress values that the material is likely to experience. A
brief description of the modelling approach relevant to the current study is discussed
here. A more detailed description of different constitutive models and their
applicability have been reviewed by Chen et al. (Chen et al. 1990).
The best known failure criterion in soil mechanics is the Coulomb criterion, which is
the first type of failure criterion that takes into consideration the effect of the
hydrostatic pressure on the strength of granular materials. This criterion states that
the resistance to failure of a material is a constant shear strength plus a friction-like
force, i.e.
= c n tan (1)
where is the shear stress, and n is the normal stress (compressive stress as a
negative quantity and tensile stress as a positive quantity), c and are the cohesion
and the angle of internal friction, respectively. The coulombs failure criterion is an
irregular hexagonal pyramid in the principal stress space. The cross-sectional shape
of this pyramid on the -plane is shown in Figure 2.30.
Figure 2.30: Drucker-Prager and Coulomb criteria on the -plane
The Drucker-Prager criterion, formulated in 1952, is major advance in the extension

of metal plasticity to soil plasticity where the influence of a hydrostatic stress
component on failure is introduced by inclusion of an additional term I1 in the von
Mises expression.
f ( I 1 , J 2 ) = I 1 + J 2 k = 0 (2)
where I1 is the first invariant of stress tensor, and J2 is the second invariant of
deviatoric stress tensor. and k are material constants. The failure surface of the
Drucker-Prager criterion in principal stress space is a circular cone, as shown in
Figure 2.31
Figure 2.31: Drucker-Prager criterion in principal stress space
The Drucker-Prager criterion can be made to match with the apex of the Coulomb
criterion for either point A or B on its -plane as shown in Figure 2.30. For Point A,
where the cone circumscribes the hexagonal pyramid (the outer cone), the two
surfaces are made to coincide along the compressive meridian (Lode angle = /3),
and the Drucker-Prager parameters and k are related to the Coulomb constant c
and by
2 sin 6c cos
= ,k = (3)
3 (3 sin ) 3 (3 sin )
While for point B (the inner cone) the two surfaces are matched along the tensile
meridian (Lode angle = 0), and will have the constants
2 sin 6c cos
= ,k = (4)
3 (3 + sin ) 3 (3 + sin )
As the material constants in ANSYS DP model are chosen to match with the
compressive meridian of the Coulomb criterion, therefore, the outer cone yield
surface is selected. The corresponding yield surface in p-q plane is shown in Figure
2.32, where p is the mean stress or hydrostatic stress and q is Mises equivalent
deviatoric stress.
q
Drucker-Prager yield criterion
6c cos
3-sin
c cos p
sin
Figure 2.32: Drucker-Prager criterion in p-q plane
The flow theory of plasticity is based on three basic assumptions: (1) the existence of
an initial yield surface; (2) the evolution of subsequent loading surfaces (hardening
rule); (3) the determination of an appropriate flow rule. For the current study, the
model is developed using Drucker-Prager material model implement in ANSYS,
therefore, the yield surface does not change with progressive yielding, hence there is
no hardening rule and the material is assumed elastic - perfectly plastic.
2.4.1.3 Contact interface modelling approach

A variety of numerical approaches has been proposed for the modelling of interface
problems. All the methods are essentially attempt to prevent the overlapping of the
finite element mesh and to give a satisfactory stress distribution over the contact
regions. As the current model is created by ANSYS, a simple frictional contact
surface is adopted for the soil-arch interface and Augmented Lagrangian method is
adopted.
2.4.2 Case study - flexible and smooth strip footings on stratum of

clay
Validation of the soil model to be applied to the soil-arch interaction FE model was
undertaken by using Zienkiewicz et al (1975) model for a shallow layer of clay as
Zienkiewicz et al (1975) analyzed a shallow layer of clay employing the Coulomb
yield condition with a non-associated flow rule as well as the associated flow rule.
The footing was assumed to be flexible and smooth so that the stresses beneath the
footing are distributed vertically and uniformly.
The footing is 10.28 ft (3.14 m) wide. The horizontal extent of the stratum is 12 ft
(7.32 m) from the footing center and the depth of the stratum is 12 ft (3.66 m). The
vertical boundary is assumed to be perfectly smooth and rigid.
The footing is re-analysed using Drucker-Prager model implemented in ANSYS

(Release 9.0). As the material constants in ANSYS DP model are chosen to match
with the compressive meridian of the Coulomb criterion, therefore, the outer cone
yield surface is selected. The plane strain finite element model using PLANE42
element is shown in Figure3. Using symmetry, only half of the model is analysed.
Figure 2.33: FE model for a shallow layer of clay
The material constants used is shown in Table 1. The effect of the weight of the soil
is neglected.
Table 2-2. Material properties for simple soil-arch model
Material constants
Youngs modulus N/mm2 207
Poissons Ratio 0.3
Cohesion N/mm2 0.069
Angle of internal friction degree 20
Both the associated and non-associated flow rule were adopted for the analyses of
the footing. The load-displacement relationship from the analyses is shown in Figure
2.34. The displacement vector using both associated flow rule and non-associated
flow rule are shown in Figure 2.35.
The collapse load using associated flow rule is around 2.5 MPa, more than twice that
of the load (152 psi or 1.05 MPa) predicted by the Coulomb criterion adopted by
Zienkiewicz et al., The load predicted by the Coulomb criterion is close to the loads
given by the Terzaghi and Prandtl solutions (1.2 MPa and 1.0 MPa).
3
Applied pressure beneath footing (N/mm2)
associated flow rule
2
non-associated flow rule
0
0 1 2 3 4 5
Displacem ent at center of footing (m m )
Figure 2.34: Load-displacement relationship for flexible footing model
Associated flow rule Non- associated flow rule

Figure 2.35: Displacement vector using associated and non-associated flow rule
As a result, the analysis with the Drucker-Prager material constants matched with the
compressive meridian of the Coulomb criterion does not agree with the well-known
Terzaghi and Prandtl solutions. Studies by others (Griffiths, 1990) also suggested
that Drucker-Prager outer cone over-predicts the strength of soil. Therefore, the
selection of the material constants using the Drucker-Prager criterion is critical.
2.4.3 Finite element models

A simple soil-arch interaction model and a full bridge model were created using
commercially available finite element package ANSYS 9.0. The simple soil-arch
interaction model is mainly used for parametric studies within service loading range.
The material for both arch and soil in the simple soil-arch model is assumed within
elastic limit and the interface between the arch and soil is characterised as a friction
contact interface. The full bridge model, however, has been created for predicting the
failure of the same corresponding experiment. In the full bridge model, the smeared
cracking approach is adopted for modelling the masonry behaviour, while the
nonlinear behaviour of the soil is simulated with a Drucker-Prager material model.
The same friction contact interface is adopted for the full bridge model.
2.4.3.1 Simple soil-arch interaction model

The geometry of simple soil-arch interaction model is shown in Figure 2.36. The arch
is of 3m span with span to rise ratio of 4:1. The thickness of the arch ring is 215mm.
The geometry of soil is control by SSSL, SLSL and h_soil.
The arch and soil were both modelled using SOLID45 (8-node solid element), while
contact element CONTACT173 and target element TARGET170 (4-node interface
elements) are used for modelling the interface between soil and arch barrel interface.
A typical meshed simple soil-arch interaction model is shown in Figure 2.37.
h_soil
5
21
750
SSSL 3000 SLSL
Figure 2.36: Geometry of simple soil-arch interaction model
Figure 2.37: FE mesh, loading and boundary condition for simple soil-arch interaction
model
The boundary condition for the simple soil-arch interaction model is shown in Figure
2.37. The arch barrel was fixed at both supports, and the plain strain condition is
assumed for the soil.
A vertical patch load (width of 219mm) was applied at the quarter span position and
the maximum applied pressure 0.3N/mm2 is determined by controlling the maximum
tensile stress in the arch barrel which is no greater than the assumed tensile strength
of the material (0.5N/mm2 based on previous experimental data for similar brickwork
arches (Melbourne et al. 2007).
The material properties for the arch and the soil are given in Table 2-3. The
coefficient of friction for the interface is assumed to be 0.7 except for the study of the
influence of this parameter (varies from 0 to 1.0).
Table 2-3. Material properties for simple soil-arch model
Material Arch barrel Soil
Youngs modulus (N/mm2) 16000 1000*
Poissons Ratio 0.2 0.2
Density (Kg/ m3) 2200 2000
2
*varies from 200 to 10000N/mm for relative stiffness studies
2.4.3.2 Full bridge model

The geometry of the full bridge model is shown in Figure 2.38, based on the
experimental data of Bridge1.
In order to study the fixity conditions of abutments, the full bridge model had two
different abutment supports: one with abutments comprised top and bottom parts, the
other with the same size but only comprised one part.
500
300
215
2165
1665
220
750
240
530
730 322 3000 325 3923
8300
Figure 2.38: Geometry of full bridge model
The arch was model using SOLID65 element with cracking and crushing capabilities,
while the soil was model using the same element but with Drucker-Prager properties.
The same interface elements as the simple soil-arch model are used for the full
bridge models. Due to the three-dimensional nature of the element SOLID65, three-
dimensional models were created. For simplicity and computing efficiency, only one
layer of elements were generated along the width of the model considering the load
was applied uniformly along the width of the bridge. A typical meshed full bridge
model is shown in Figure 2.39.
Figure 2.39: FE mesh of full bridge model
The boundary condition for the full bridge model is similar to that of simple soil-arch
interaction model except the soil was constrained by rigid tank which is fixed in all
direction and the lower section of the abutments are fixed at the bottom face. A
vertical patch load (width of 219mm) of the same size and same location as the full-
scale bridge test was applied incrementally
Table 2-4. Material properties for full bridge model
Arch barrel soil
2
Youngs modulus N/mm 16000 1000
Poissons Ratio 0.2 0.2
Density Kg/ m3 2200 1910
Uniaxial tensile cracking stress N/mm2 0.48
Uniaxial compressive stress N/ mm2 24
Cohesion N/ mm2 0.0224
Internal angle of friction degree 46.4
2.4.4 Finite element results and discussion
2.4.4.1 Simple soil-arch interaction model

From the simple soil-arch interaction model, the following studies have been carried
out.
The influence of the geometry of the soil
Figure 2.40shows the influence of SSSL, SLSL and the depth of fill above the crown
(h_soil) on the maximum tensile principal stress in the intrados of the arch barrel
(S1max), and the maximum von Mises stress (SEQV) in the soil.
The influence of the horizontal extent of the soil on both side of the supports on
stress and displacement is generally insignificant especially when they are over 1m.
This indicates that at this level of loading, only the soil immediately surrounding the
arch barrel has interacted with the arch barrel. The rest of the studies are based on
SSSL = 1m and SLSL = 2m.
The influence of the depth of fill above the crown on the maximum tensile principal
stress in the arch barrel is significant. When the fill cover approaches 1.2m the model
predicts that it is unlikely for the arch barrel to crack before the yielding of the soil. It
also suggests that when there is sufficient fill cover, the load can be dispersed
directly to the base of the soil.
0.6 0.6
0.6
0.5 0.5
0.5
Stress (N/mm2)
S1max
Stress (N/mm2)
Stress (N/mm2)
0.4 S1max 0.4 S1max 0.4

SEQV SEQV
0.3 SEQV
0.3 0.3
0.2 0.2 0.2
0.1 0.1 0.1

0 0
0
0 1000 2000 3000 4000 0 1000 2000 3000 4000 5000
0 500 1000 1500 2000 2500
SSSL (m m ) SLSL (m m)
h_soil (m m )
Figure 2.40: Influence of the geometry of the soil
The influence of relative stiffness of arch barrel to the soil

The influence of the relative stiffness of arch barrel to the soil is studied by keeping
the Youngs modulus of the arch constant (E_arch=16000N/mm2) and varying the
Youngs modulus of the soil (E_soil) from 200 to 10000N/mm2. The influence on
stress in arch barrel and soil is shown in Figure 2.41.
0.8
0.7
0.6
Stress (N/mm2)
0.5
S1max
0.4 SEQV
0.3
0.2
0.1
0
0 20 40 60 80
E_arch/E_soil
Figure 2.41: Influence of the relative stiffness between arch barrel and soil
The influence of relative stiffness is clearly very significant not only on the value of
the maximum stresses but also on the positions of where these stresses occur. For
most case (except for the E_ratio < 4), the maximum tensile stress occurs in the
intrados of the arch barrel indicating that if the same yield criteria is used for both soil
and arch, the yielding or cracking will initiated from the arch barrel intrados.
The influence of contact stiffness on soil arch interface
Special care should be taken when selecting the values of contact stiffness between
arch barrel and soil interface. Normal contact stiffness is used to enforce compatibility
between the contact surfaces. If the contact stiffness factor (KFN) is too small, the
amount of penetration of contact surface into target surface may be too great resulted
in fictitious soft interface and the solution can be incorrect. On the other hand, if the
stiffness is too big, the determination of the true contact status normally requires
more iterations, and in some cases, convergence difficulties are inevitable.
The study has shown that the normal contact stiffness factor of 0.1 gives reasonable
stress levels in the arch barrel and prevent overlapping between interfaces
(penetration smaller than 0.1mm), it also leads to efficient solutions in terms of the
number of iterations.
The influence of mesh density
The art of using the finite element method lies in an appropriate mesh density to
solve a problem. If the mesh is too coarse then the inherent element approximations
will not allow a correct solution to be obtained. Alternatively, if the mesh is too fine the
cost of the analysis can be out of proportion to the results obtained. It is therefore
important to use a sufficiently refined mesh to ensure that the results from FE
simulation are adequate.
As far as the deflection is concerned even a coarse mesh will result in easy
convergence. For the convergence of stress, at least four elements are need across
the thickness of the arch barrel. For the maximum principal stress at an integration
point, the convergence is much slower and the difference between stress from
integration point and stress extrapolating from integration point always exists even
with a finer mesh (more than eight elements across the thickness). Mesh sensitivity
will, therefore, exist when cracking is based on the stress at integration points. The
rest of the studies are based on four elements across the thickness of the arch barrel.
The influence of the coefficient of friction on the interface
The influence of the coefficient of friction (MU) between the arch barrel and soil
interface is studied by varying the MU from 0 (frictionless contact) to 1.0 (rough
contact). The influence on stress in arch barrel and soil is shown in Figure 2.42. The
influence on the stress level is not very significant considering the contact interface
has changed from frictionless contact to relatively rough contact.
0.6
0.5
Stress (N/mm2)
0.4
S1max
0.3
SEQV
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2
MU
Figure 2.42: Influence of the coefficient of friction on the interface
The influence of abutment fixity and soil boundary conditions

The influence of the boundary condition on the bottom face of the soil is studied by
changing from perfectly smooth and rigid to perfectly rough and rigid. The influence
on the maximum displacement and the stress level in both arch barrel and soil is
shown in .
Table 2-5. The influence of the soil boundary condition within elastic range of loading
is not as significant as the abutment fixity of the arch barrel, see Table 2-6.
Table 2-5. influence of soil boundary condition
S1max USUM SEQV
Soil bottom face boundary conditions N/mm2 mm N/mm2
Perfectly smooth and rigid 0.4863 0.2389 0.2029
Perfectly rough and rigid 0.4762 0.2336 0.2038
Table 2-6. influence of abutment fixity

S1max USUM SEQV
abutment boundary conditions N/mm2 mm N/mm2
Both sides fixed 0.4863 0.2389 0.2029
Both sides free to move horizontally 1.327 0.4104 0.2094
Although the maximum horizontal movement on both side is less than 0.2mm at the
current load level, the maximum principal stress in the arch barrel has nearly tripled,
and the position of the stress has moved from quarter span intrados to the extrados
of left-hand abutment, which indicates that the crack will start from here rather than
the quarter span intrados.
2.4.4.2 Full bridge model

Figure 2.43 and Figure 2.44 show the principal stress distribution in the full bridge
model with two different abutment supports: one with abutments comprised top and
bottom parts, the other with the same size but only comprised one part.
At the same load level, the stress in the arch barrel with top and bottom parts
abutment support was higher than that with one part abutment support. Therefore,
the load causing the first cracks to occur would be lower. The load for the occurrence
of the first crack was 54kN/m for the abutment comprised two parts (top and bottom)
and 59kN/m for the corresponding bridge with abutment comprised one part.
Figure 2.45 show the displacement vector from the FE full bridge models with two
different abutment supports. Clearly with moveable abutment (abutment comprised
top and bottom parts), more soil was mobilised.
Figure 2.43: Principal stress (S1) distribution in the arch barrel and the bridge
(abutment comprised one part) (P=59kN/m)
Figure 2.44: Principal stress (S1) distribution in the arch barrel and the bridge
(abutment comprised two parts) (P=59kN/m)
Figure 2.45: Displacement vector (P=59kN/m)
The bridges eventually failed by the formation of hinged mechanism plus the sliding
between top and bottom parts of the abutments. All these essential features are
similar to the large-scale bridge test.
The study has shown that the maximum load predict from the FE model is very
sensitive to the material properties such as cohesion, angle of friction, dilatancy,
tensile strength of the brickwork, density and relative stiffness between brickwork and
soil. Therefore, interpretation of these parameters from lab tests is critical.
Figure 2.46 shows an example of the influence of relative stiffness of the arch barrel
to the backfill. With relatively stiffer backfill, the bridge can carry much higher load
than that with softer backfill.
Figure 2.46: Cracks in the arch barrel for different stiffness of backfill (P=164kN/m)
2.5 Conclusions
From the experimental programme, the following conclusions have been drawn:
Tests on pilot crushed limestone and clay filled bridges have been performed.
All the bridges tested failed by hinged mechanisms, although some abutment
movement was recorded in the tests;
The limestone filled arch bridges proved capable of carrying more load than
clay filled bridge. This demonstrates the importance of the fill type and
indicates that performing intrusive investigation to ascertain fill type is likely to
be worthwhile in the case of bridges with borderline load carrying capacity;
The tests indicated that the test rig performed as designed, with minimal frame
deflection (thereby giving effectively plane strain conditions);
Inclusion of large windows along one side of the test chamber permitted the
acquisition of good quality particle image velocimetry (PIV) data which is being
used to help better understand the nature of the soil-arch interaction.
The following conclusions can be drawn from the current finite element analyses:
The relative stiffness of the arch and the soil is a very important factor on soil-
arch interaction;
The fixity of the abutments affect the load at which the 1st crack occurs and
will in turn affect the load redistribution in the bridge;
The full bridge model can predict the essential feature observed in full scale
experiments;
The load carrying capacity from the FE model is very sensitive to the material
properties like density, cohesion of soil, the internal angle of friction, dilatancy
and the tensile strength of the brickwork. Therefore, interpretation of these
parameters from lab tests is critical.
3. Acknowledgement
The authors wish to acknowledge the financial and technical support provided by
Network Rail and Essex County Council throughout the project.
The soil-arch interaction project is in collaboration with Sheffield University and the
authors acknowledge the financial support provided by Essex County Council and
Network Rail.
4. References
Bazant, Z., Planas, J. (1997). Fracture and Size Effects in Concrete and Other
Quasi-brittle Materials, CRC Press, USA.
Boothby, T. E. and B. J. Roberts (2001). "Transverse behaviour of masonry arch
bridges." The Structure Engineer Volume 79(9): 21-26.
Burroughs, P., T. Hughes, S. Hee, and M. Davies (2002). "Passive pressure
development in masonry arch bridges." Proc. ICE, Structures & Buildings 152(4):
331-339.
Chen, W. F. (1985). Constitutive Relations for Concrete, Rock and Soils: Discusser's
Report. Mechanics of Geomaterials. Z. Bazant, John Wiley & Sons Ltd.: 65-73.
Chen, W. F. a. E. M. (1990). Non-linear Analysis in Soil Mechanics - Theory and
Implementation. The Netherlands, Elsevier Science Publishers B.V.
Edgell, G. (2005). Testing of Ceramics in Construction. Whittles Publishing,
Caithness, ISBN 1-870325-43-5.
Fairfield, C. a. D. P. (1994). "Model tests to determine the effect of fill on buried
arches." Proc. ICE Structures and Buildings 104(4): 471-482.
Fang, Y., Chen, T., Holtz, R. and Lee, W. (2004). "Reduction of boundary friction in
model tests." Geotechnical Testing Journal 27(1): 3-12.
Fanning, P. J., Boothby, T. E. (2001). Non-linear three dimensional simulations of
service load tests on a 32m stone arch bridge in Ireland. Arch'01 Third international
arch bridges conference, Paris.
Garrity, S. W. and Toropova, I. L. (2001). A finite element study of a single span
masonry arch bridge with near surface reinforcement. Arch'01 Third international
arch bridges conference, Paris.
Gilbert, M. Smith, C.C., Melbourne. C and Wang. J . (2006). An experimental study of
soil-arch interaction in masonry bridges. IABMAS'06 - Third International Conference
on Bridge Maintenance, Safety and Management, Portugal.
Griffiths, D. V. (1990). "Failure criterion interpretation based on Mohr-Coulomb
friction." J. Geotech. Engng. ASCE(116): 986-999.
Heyman, J. (1966). The stone skeleton. Structural Engineering of Masonry
Architecture. University of Cambridge, ISBN-13 9780521629638, ISBN-10
0521629632.
Heyman, J. (1996). Arches, Vaults and Buttresses, Masonry Structures and their
Engineering Valorium. ISBN 0860785971.
Highway structures: Design (Substructures and Special Structures), Materials.
Special Structures. Unreinforced Masonry Arch Bridges. DMRB Volume 2 Section 2
Part 14 (BD 91/04), Highways Agency, The Stationary Office, London.
Highway structures: Inspection and maintenance, Inspection, Advice Notes on the
Non-destructive Testing of Highway Structures. (2004). DMRB Volume 3 Section 1
Part 7 (BA 86/04), Highways Agency, The Stationary Office, London, Highways
Agency Advice Note BA86.
Hogg, V. (1997). Effects of Repeated Loading on Masonry Arch Bridges and
Implications for Serviceability Limit State. PhD Thesis, University of Nottingham.
Hughes, T.G., Davies, M.C.R., Taunton, P.R. (1998). The Small Scale Modelling of
Masonry Arch Bridges Using a Centrifuge. Proc. Inst. Civ. Engrs., Structures &
Building Journal, 128, pp. 49-58.
Lange, D.A., Jennings, H.M., Shah, S.P. (1993). Relationship Between Fracture
Surface Roughness and Fracture Behavior of Cement Paste and Mortar. Journal of
American Ceramic Society, 79, (3), 589-597.
Lourenco, P. B. (1996). Computational Strategies for Masonry Structures, PhD
thesis, Delft University of Technology.
Melbourne, C. and Gilbert, M. (1995). "The behaviour of multi-ring brickwork arch
bridges." The Structural Engineer 73(3): 39-47.
Melbourne, C., Tomor, A.K., Wang, J., (2005). Modes of Failure of Multi-ring
Masonry Arches under Fatigue Loading. Proc. Bridge Management Five
Conference, UK, pp 476-483.
Melbourne, C., Tomor, A.K., (2006). A New Assessment Method for Masonry Arch
Bridges. Proc. IABMAS06 Conference, Porto, Portugal.
Melbourne, C. Wang, J. and Tomor, A. (2007). "New Masonry Arch Bridge
Assessment Strategy (SMART)." Proc. ICE, Bridge Engineering 160: 81-87.
Tomor, A.K., Melbourne, C. (2007). Monitoring Masonry Arch Bridge Response to
Traffic Loading Using Acoustic Emission Techniques. Proc. ARCH07 Conference,
Madeira.
Tomor, A.K., Melbourne, C., (2007). Condition Monitoring of Masonry Arch Bridges
Using Acoustic Emission Techniques, Structural Engineering International (IABSE),
2/2007.
Page, J. (1993). Masonry Arch Bridges. London, HMSO.
Page, J. (1995). (1995). Load Tests to Collapse on Masonry Arch Bridges. Thomas
Telford, London, pp. 289-298.
Pippard, A J.S. (1948). The Approximate Estimation of Safe Loads on Masonry Arch
Bridges. Civil Engineer in War, 1, 365-372, ICE, London.
Pippard, A.J.S, Chitty, L. (1951). A Study of the Voussior Arch. National Building
Series, Research Paper 11, London.
Tao, H. (2001). The behaviour of open spandrel brickwork masonry arch bridges.
PhD thesis, University of Salford.
Wang, J. (2004). Three dimensional Behaviour of Masonry Arches. PhD thesis,
University of Salford.
Wang. J., Smith, C.C., Gilbert, M. and Melbourne. C (2006). Load test to collapse of
back-filled brickwork masonry arch bridges. Data Report: Salford-Soil-Arch-0001 and
Salford-Soil-Arch-0002.
White, D. a. T., W. (2002). GeoPIV particle image velocimetry (PIV) software for use
in geotechnical testing. Technical Report 322. Cambridge, Geotechnics Group,
Cambridge University Engineering Department.
Willam, K. J. and E. D. Warnke (1975). Constitutive Model for the Triaxial Behavior of
Concrete. Proceedings, International Association for Bridge and Structural
Engineering, ISMES. Bergamo, Italy, ISMES. Vol. 19: 174.
Zienkiewicz, O. C., Humpheson, C. and Lewis, R. W. (1975). "Associated and non-
associated visco-plasticity and plasticity in soil mechanics." Geotechnique 25(4): 671-
689.
Numerical Analyses of Load Distribution and
Deflections in Railway Bridge Transition Zones due to
Passing Trains
PRIORITY 6
INTEGRATED PROJECT

Author/s: G. Holm, M. Smith, P.E. Bengtsson, SGI
Revision date:

Dissemination Level
PU Public X
PREFACE ................................................................................................................................1
SUMMARY...............................................................................................................................2
1 INTRODUCTION...............................................................................................................4
2 BACKGROUND................................................................................................................5
2.1 Railway Traffic on Soft Soil Sites ...............................................................................5
2.2 Railway Traffic in Bridge Transition Zones ................................................................6
3 NUMERICAL VERIFICATION STUDY .............................................................................9
3.1 Embankment and Instrumentation .............................................................................9
3.2 FLAC3D Model ........................................................................................................10
3.3 Results of FLAC3D Analyses...................................................................................12
3.4 Closing Remarks......................................................................................................14
4 NUMERICAL ANALYSES WITH CONCRETE BRIDGE................................................15
4.1 FLAC3D Model and Analyses..................................................................................15
4.2 Results of Parametric Analyses ...............................................................................17
4.2.1 Train Axle Load.................................................................................................17
4.2.2 Train Velocity ....................................................................................................21
4.2.3 Material Property Values of Ballast, Subballast, and Embankment Fill ............24
4.3 Summary and Conclusions ......................................................................................30
5 NUMERICAL ANALYSES WITH MASONRY ARCH BRIDGE ......................................31
5.1 FLAC3D Model and Analyses..................................................................................31
5.2 Results of Analyses .................................................................................................32
5.3 Summary and Conclusions ......................................................................................38
6 LIMITATIONS .................................................................................................................40
7 CONCLUSIONS..............................................................................................................41
8 REFERENCES................................................................................................................42
PREFACE
For the assessment of existing railway bridges it is important to have a holistic approach that
takes into account the bridge, foundation and the embankment transition zones. The load
distribution and deflections in the transition zones due to passing trains, which will be faster
and heavier in the future, should be assessed with respect to the influence on the bridge.
The main development achieved in this work package of the Sustainable Bridges project are
numerical analyses used to evaluate the effect of train velocity and load, as well as the stiff-
ness of the ballast, subballast, and embankment material on the stresses and deflections in
the transition zones. Three dimensional dynamic analyses were performed to study two
bridges representative of existing European railway bridges, one concrete bridge and one
masonry arch bridge. No laboratory or field tests have been performed. However a numeri-
cal verification study was performed using the measured data at a full-scale instrumented
railway embankment in Finland (Kolisoja and Mrkel, 2001). This study shows good agree-
ment between the 3D numerical analyses and pressure cell data. The obtained results indi-
cate that 3D numerical analyses can be a useful tool when evaluating the load distribution
and deflections in transition zones due to passing trains. The findings of the numerical
analyses should be verified against laboratory scaled tests and/or full-scale field tests.
The technical development and this report have been financed by the Sustainable Bridge
project.
SUMMARY
Many railway lines in Europe are 60 to 100 years old, and are not designed in accordance with
modern railway traffic. As identified by many railway authorities, faster and heavier modern
trains are causing problems at existing railway bridges. Bridge ends are one of the most com-
plicated parts of a railway track since, at these locations, the rail is subjected to concentrated
stresses, which are due to dynamic and cyclic loads caused by the passage of trains and
bridge deflections. The extent of the problems that develop within the transition zone is de-
pendent upon the train load and velocity, and subgrade stiffness. Unlike train traffic on soft
sites, little has been published about the effect of train traffic on the load distribution and de-
flections in the bridge transition zone.
The scope of this work included performing numerical analyses to evaluate the load distribu-
tion and deflections in bridge transition zones due to the passing of trains. No field or labora-
tory tests were performed. This document is a background document, intended to provide
insight into the general behavior of bridge systems under the passing of high speed trains.
No design methods or recommendations are presented herein.
Numerical analyses consisted of three-dimensional dynamic analyses. The numerical mod-

elling methods were verified with data from pressure cells installed at a test embankment by
railway authorities in Finland. The Finnish test embankment did not feature a bridge struc-
ture; however, the three-dimensional dynamic numerical analyses provided a good fit to the
pressure cell data from the test embankment.
Using a bridge geometry representative of a typical European concrete bridge, a numerical

parametric study was performed to study the effects of 1.) the train load, 2.) the train velocity,
and 3.) the stiffness of the ballast, subballast and embankment fill.
Key findings from the numerical analyses include:
The greatest increases in vertical and horizontal stresses are concentrated within the
upper 1.0 to 1.2 m of the system. Vertical and horizontal stresses caused by the pass-
ing of a train decrease with depth.
As train axle loads increase, the vertical and horizontal stresses, and vertical deflec-
tions beneath the train increase.
As train velocity increases, the horizontal stresses and vertical deflections increase,
however, they do not increase linearly with increasing velocity. As the velocity of the
train approaches the Rayleigh wave velocity of the embankment system, significant
horizontal stresses and vertical deflections develop.
As the stiffness of the ballast and/or subballast materials increase, net horizontal
stresses exerted on the bridge decrease. Horizontal stresses in the embankment fill
are relatively small, but increase when a soft fill is used.
When a soft embankment fill is used, vertical deflections increase substantially, and
these deflections are observed to occur 3 m behind the train, as if there is a stern
wave behind the train.
Numerical analyses were also performed using a bridge geometry representative of Euro-
pean masonry arch bridges. It can be assumed that the factors that affect the stress distribu-
tion and deflections for the concrete bridge structure (as described above) will have the same
effect for the masonry arch bridge structure. However, the geometry of the arch bridge is
different than the concrete bridge in two significant ways: (1) for masonry arch bridges there
is a gradual change in stiffness, while for a concrete bridge the change is more abrupt, and
(2) for the masonry arch structure, the ballast and subballast are confined within a relatively
small volume. Due to the latter factor, additional vertical and horizontal stresses induced by
the train are distributed immediately to the arch structure. Since there is less soil above the
crown of the arch to dissipate stresses, the crown of the arch will experience the greatest
increase in vertical stress under the passing of a train.
1 INTRODUCTION
Many railway lines in Europe are 60 to 100 years old, and are not designed in accordance
with modern railway traffic. Due to faster and heavier modern trains, existing railway bridges
are experiencing problems. These issues have an adverse effect on the safety, reliability,
and economy of the railway line, and therefore, many existing bridge systems require up-
grading. Engineers are faced with the task of assessing the performance of existing bridges,
and, if necessary, designing the strengthening or repair systems.
tory tests were performed. Numerical analyses consisted of three-dimensional dynamic
analyses. The numerical models are representative of two common European bridge types,
a concrete and a masonry arch bridge.
2 BACKGROUND
2.1 Railway Traffic on Soft Soil Sites

At low train velocities, the ground response from a moving load is essentially quasi-static
(Kaynia et al. 2000, Chen et al. 2005); however, vertical movements increase substantially
as the train velocity approaches the Rayleigh-wave velocity of the soil (Adolfsson et al.
1999b, Kaynia et al. 2000). The velocity of the train that equals the Rayleigh-wave velocity
of the soil is commonly referred to as the critical velocity. Close to the critical train velocity,
the dynamic stresses and strains in the embankment and subsoil materials are high and ma-
terials behave non-linearly (Madshus et al. 2004). As the train velocity approaches the critical
velocity, deformations and vibrations within the soil increase significantly. Therefore, sites
with low Rayleigh-wave, or shear-wave, velocities, such as sites with soft soils, are particu-
larly susceptible to excessive deformations and vibrations from high-speed trains. Chen et
al. (2005) also note that stresses induced by the train also increase with increasing train ve-
locity, and, similarly, their values will be considerable as the train velocity approaches the
Rayleigh-wave velocity of the soil.
Kopf and Adam (2005) performed a parametric study using an analytical model they devel-
oped. Kopf and Adam (2005) observed that the critical velocity of the system decreases with
increasing train load. They also observed that in the static case, the maximum deflection
occurs beneath the applied load; whereas in the dynamic case, there is a time lag between
the maximum applied force and the maximum deflection. As the train velocity increases, the
wave length of the bow wave in front of the load becomes shorter while the rear wave be-
hind the load becomes longer (Kopf and Adam 2005).
Kopf and Adam (2005) also present the results of field tests on a high speed railway line be-
tween Vienna and Salzburg. Pressure cells were placed beneath the railway sleepers, and a
train was passed over the tracks at a range of velocities up to 230 km/hr. The pressure cell
measurements revealed that the maximum dynamic pressure beneath the train increased
with increasing train velocity.
Katzenbach and Ittershagen (2005) describe field tests that were performed to evaluate the
dynamic soil-structure interaction of railway lines on soft soil sites. Stress measurements were
taken within the soil at depths of 0.4 m beneath the sleepers under the passing of passenger
(axle load = 100 kN) and freight (axle load = 225 kN) trains. Katzenbach and Ittershagen
(2005) observed that each crossing of a train axle caused a single stress amplitude. Stress
amplitudes were less for the passenger train, which had smaller axle loads, than the freight
train. Measured particle velocities decreased with depth, and increased with increasing train
velocity (Katzenbach and Ittershagen 2005).
Adolfsson et al. (1999a) and Adolfsson et al. (1999b) describe a project funded by Banverket
to evaluate deflections and accelerations in soft soil due to the passing of trains with veloci-
ties up to 204 km/hr. The project site is located in Ledsgrd, approximately 25 km south of
Gothenburg, Sweden. An X2000 train set was used and the train velocity was varied from
zero (static) to 204 km/hr. Axle loads were on the order of 120 to 190 kN. The embankment
at the project site consisted of 0.5 m of ballast underlain by 0.9 m of gravel. Below the em-
bankment, the soil conditions consisted of 3 m of gyttja (organic clay) underlain by more than
50 m of soft clay. The shear wave velocity of the soft soil increased from 45 m/s to 90 m/s at
a depth of 14 m (Adolfsson et al. 1999a). The railway and embankment were instrumented
with displacement transducers and accelerometers. There was no bridge at the project site.
Based on the measurements at the Ledsgrd site, Adolfsson et al. (1999a) observed that, at
low train velocities, the vertical deflections were static and occurred in the upper 5 m of the
soil. At train velocities 200 km/hr or greater, the deflections were approximately three times
larger and occurred to depths up to 10 m. Adolfsson et al. (1999b) also observed that the
maximum downward deflections usually occurred below the bogie axles of the heaviest
wagon of the train set, which is the engine wagon for the X2000 train. Adolfsson et al.
(1999b) point out that, for higher train velocities, a traveling wave in the embankment is de-
tected by the instrumentation behind the train, much like a stern wave behind the ship.
Numerical analyses described by Adolfsson et al. (1999b) confirm the tail wave phenomena.
The numerical analyses show the shifting from a static-like deflection pattern of the ground
surface at low train velocities to a plough-shaped dynamic pattern at high train velocities
(Adolfsson et al. 1999b).
Furthermore, the high vibration levels detected at the Ledsgrd site were detected under the
passing of the X2000 train but not under the passing of cargo trains. Therefore, it was de-
duced that the vibration problems were related to the high velocity of the X2000 train, and not
the train axle loads (Adolfsson et al. 1999a).
Lundqvist and Dahlberg (2004) note that track settlements are often associated with the dy-
namic loading of high-speed train traffic, and that high-speed railway traffic has caused track
settlements, even if the high-speed trains have relatively low axle loads. However, based an
analytical model presented by Hunt (1997), the total magnitude of the track settlement is
governed largely by the magnitude of the axle loads, although some local effects are con-
trolled by dynamic loads. Hunt (1997) noted that for train velocities up to 350 km/hr, the ve-
locity of the wave propagation in most tracks is very much greater than the train velocity.
The response of the track system is highly dependent upon the Rayleigh-wave velocity of the
soil and structural components. As a rule of thumb, Madshus et al. (2004) recommend that
track systems have a critical speed of at least 1.7 times the operating maximum speed of the
train. They note that there are basically two ways to increase the critical speed of the track
structure-ground response: (1) by improving the stiffness and strength of the ground be-
neath the track structure, and (2) by increasing the longitudinal bending stiffness and thus
increasing the bending wave speed along the track itself.
2.2 Railway Traffic in Bridge Transition Zones

Bridge ends are one of the most complicated parts of a railway track since, at these locations,
the rail is subjected to concentrated stresses, which are due to dynamic and cyclic loads
caused by the passage of trains and bridge deflections (ERRI 1999). The extent of the prob-
lems that develop within the transition zone is dependent upon the train load and velocity,
subgrade stiffness, and the length of the transition zone.
Many of the problems that occur within the transition zones are associated with modern train
traffic, which is heavier and faster than the traffic for which the bridges were designed. Due to
faster and heavier modern trains, existing railway bridges are experiencing problems, such as
an increase of differential settlements within the bridge transition zone, and an increase of
loads on existing bridge structures. An illustration of problems that can develop in bridge tran-
sition zones is shown schematically in Figure 2.1.
Train
Ballast Layer
Bridge Structure Increased vertical deformations
Increased vertical and
horizontal stresses behind
the bridge
Embankment
Subsoil
Figure 2.1. Illustration of problems that develop at bridge transition zones

as train loads and velocities increase
ERRI (1999) developed a State of the Art report based on the responses of 14 participating
European railways to a questionnaire relating railway track, bridges, and earthwork. A few of
the important conclusions made by ERRI (1999) include:
There is insufficient coordination between track, bridge and geotechnical engineers.
In many countries, the transition zones of existing lines give rise to more problems
and require more maintenance than do new lines.
There is an urgent need to develop methods for upgrading existing transitions zones
which can be implemented with minimum disturbance to traffic.
ERRI (1999) states that maintenance within the transition zone is required five times more
than for other track locations. It is unknown how much of this required maintenance is al-
ready attributed to heavier and faster modern railway traffic; however, it can be expected that
as train loads and velocities increase, the amount of required maintenance will also increase.
Settlements have been identified as the major contributory factor in deterioration of the track
geometry in bridge transition zones (ERRI 1999). Settlements may occur in the subsoil and/or
in the bridge backfill material. Li and Davis (2005) identified three major causes of settlement
within the transition zones:
1. A large difference in stiffness between the portion of the track on the bridge and the por-
tion in the transition zone can lead to uneven deflections;
2. Unlike the bridge, which is supported on stiff foundations, the transition zone is sup-
ported on the subsoil and inherently settles more than the bridge;
3. Settlements develop due to poor embankment materials, inadequate compaction, and
poor drainage conditions.
These problems have an adverse effect on the safety, reliability, and economy of the railway
line. Differential settlements that develop within the transition zone have several impacts on
the railway operations, such as required repeated maintenance work and restrictions on train
velocities. As soon as differential settlements begin to develop, the variations of the dynamic
train/track forces increase and this speeds up the track deterioration process (Lundqvist and
Dahlberg 2004).
Li and Wong (2005) investigated four bridge sites in Kansas, USA. The bridges are precast
concrete ballasted deck bridges and are 9 to 54 m long, with 8.4 m spans typical. All four
bridges are located on the same track route that sees a significant amount of traffic, with
70% of the traffic having axle loads of 36 tons. A discussion of the soil conditions is provided
by the authors. Li and Wong (2005) observed that the tracks in the transition zone experi-
enced more degradation than the tracks on the bridge or the tracks far away from the bridge.
Li and Wong (2005) also concluded that, for the four sites investigated, the resulting differen-
tial settlement within the transition zone came mostly from the ballast and subballast layers,
with some additional contribution from the underlying soil layers. The authors note that reme-
dies intended to strengthen the subgrade were not as effective because they did not com-
pletely address the stiffness issues associated with the bridges and the transition zones. Li
and Wong (2005) recommend that mitigation techniques such as rubber pads under the con-
crete sleepers or rubber mats on the concrete bridges be used to reduce track stiffness on
bridges, to increase damping characteristics, and to reduce impacts between the ballast parti-
cles and concrete sleepers or bridge surfaces.
Olofsson and Hakami (2000) performed three-dimensional numerical analyses to evaluate the
interaction between the train track, bridge abutment, and backfill in the transition zone. Based
on the analyses performed by Olofsson and Hakami (2000), the stiffness of the embankment
fill had a significant effect on the deflections within the transition zone. As the stiffness of the
backfill decreased, differential deflections increased. However, calculated vertical deflections
on the bridge platform were negligible.
As identified by many railway authorities (ERRI 1999, Li and Wong 2005), faster and heavier
modern trains are causing problems at existing railway bridges. However, unlike train traffic
on soft sites, less has been published about the effect of train traffic on the load distribution
and deflections in the bridge transition zone.
3 NUMERICAL VERIFICATION STUDY

Kolisoja and Mrkel (2001) present the results of field tests performed at a full-scale instru-
mented railway embankment. The field tests were performed as part of a research project by
the Finnish Rail Administration to investigate the possibilities of increasing the maximum al-
lowable train axle load to 250 or 300 kN. In addition to the field tests, Kolisoja et al. (2000)
performed numerical analyses of the test embankment using a two-dimensional multi-layer
linear elastic model. An abbreviated discussion of the field tests and numerical analyses is
presented in English by Kolisoja and Mrkel (2001); detailed discussions of the field tests and
analyses are provided in Finnish by Kolisoja et al. (2000).
The field tests were analyzed to verify the use of FLAC3D models to evaluate the load transfer
behaviour of embankments subjected to moving train loads. Results from the FLAC3D analy-
ses were compared to (1) the published results of the field measurements, and (2) the results
of two-dimensional linear elastic analyses performed by Kolisoja et al. (2000) of their own field
tests. A discussion of the behaviour observed by Kolisoja and Mrkel (2001) during the field
tests is also provided.
3.1 Embankment and Instrumentation

Field tests were performed on a two track railway embankment on the railway line between
Koria and Kouvola, Finland (Kolisoja and Mrkel 2001). A schematic cross-section of the
railway embankment is shown in Figure 3.1. In general, the embankment consists of 1 m of
ballast material overlying about 0.8 m of sand; an old layer of gravel about 0.4 m thick lies be-
low the sand layer. Beneath the gravel layer, soft clay extends to a depth of about 22 meters.
Due to settlement, the embankment was built up over a period of about 100 years, and as a
result, the embankment was not built according to any existing regulations (Kolisoja and
Mrkel 2001).
Figure 3.1. Schematic cross section of embankment (from Kolisoja and Mkel 2001)
Instrumentation consisted of 16 strain transducers and four pressure cells. The pressure cells
were installed at depths of 0.5 and 1.0 m below the bottom of the sleepers, and the strain
transducers were installed at depths of 0.5, 1.0 and 1.8 m. Measurements were taken over a
period of 44 hours during July 1999. Normal train traffic was maintained, and in addition, a
test train with four axle loads of exactly 250 kN passed over the instrumented site ten times at
velocities of 40 to 100 km/hr.
3.2 FLAC3D Model

Numerical analyses of the Finnish instrumented railway embankment were performed using
the finite difference FLAC3D (Fast Lagrangrian Analysis of Continua in 3 Dimensions) com-
puter program with the dynamic mode (Itasca 2005). The model geometry was chosen based
upon descriptions of the railway embankment provided by Kolisoja and Mrkel (2001) and
Kolisoja et al. (2000). The profile of the FLAC3D model is shown in Figure 3.2, where it can
be seen that the profile consists of an upper ballast layer 1.0 m thick, which is underlain by a
layer of sand 0.8 m thick, which in turn is underlain by a layer of gravel 0.4 m thick. A soft clay
layer with a thickness of 4.65 m underlies the embankment material. A stiff, non-yielding,
bearing layer is assumed to exist beneath the soft clay in the model.
8.2 m 3.3 m
C
L 2.5 m
Ballast 1.0 m
Sand 0.8 m 0.4 m
Gravel
Clay
4.65 m
30.0 m
Figure 3.2. Cross-section of FLAC3D model
The model is 30.0 m wide and 25.0 m long. Zone sizes are on average 0.25 x 0.27 x 0.26 m,
and a total of 129,200 zones were used. Previous mesh refinement studies have shown that
this mesh refinement was sufficient to reach numerical convergence. In FLAC3D, the spatial
element size should be smaller than approximately one-tenth to one-eighth of the wavelength
associated with the highest frequency component of the input wave (Itasca 2005). For the
analyses presented herein, the wavelength of the input wave is equal to 7.25 m, which corre-
sponds to the length of the simplified load associated with one bogie.
The approximated load representing the load beneath two wheels of one bogie is schemati-
cally shown in Figure 3.3. Numerical problems can develop when attempting to apply the
moving train load as successive point loads on structural elements (i.e. rails and sleepers);
therefore, the train load was input as a cosine-wave. The cosine-wave was applied succes-
sively over each zone of the model. The duration of the cosine-wave is equal to the length of
one bogie (7.25 m) divided by the velocity of the train. The width and magnitude of the wave
are determined based upon the theory of a loaded beam on a Winkler elastic bed, which
represents the distribution of the train load to the system by the sleepers. Granted, this load
representation is a simplification but it has been found to reasonably represent the actual
train load in analysing the high-speed phenomenon by other researchers (e.g. Andrasson
2002). This section describes further verification of the FLAC3D modelling procedure
against field data.
2.9 m
Figure 3.3. Approximation of the load from one bogie
The initial train load was calculated based upon an axle load of 250 kN. Axle loads of 125,
225, and 330 kN were also considered in the study. For all of the FLAC3D analyses in this
verification study, a train velocity of 50 km/hr was used.
The material property values used in the numerical analyses are listed in Table 3.1. A linear-
elastic perfectly plastic model with a Mohr-Coulomb failure criterion was used to represent
the ballast, sand, gravel, and clay materials. The elastic modulus of the ballast material de-
creases with depth. In the FLAC3D model, the ballast layer was divided into four sublayers
with values of elastic modulus equal to 400, 250, 175, and 155 (MPa), respectively. These
values of elastic modulus are equal to the values used by Kolisoja et al. (2000) in their two-
dimensional analyses.
Table 3.1. Material property values used in the analyses (from Kolisoja et al. 2000)
Moist Friction
Poisson's Elastic Modulus Cohesion
Material Density Angle
Ratio (MPa) (kPa)
(kN/m3) (deg)
Decreases from 400 to
Ballast 22 0.3 40 0.1
155 with depth
Sand 18 0.3 145 33 0.1
Gravel 20 0.3 160 35 0.1
Clay 16 0.3 40 20 10
No material damping was used. Additionally, no structural elements were used to model the
rail and sleepers. It was desired to model the train dynamically, and since the timestep in the
dynamic mode is determined by the largest material stiffness, using structural elements
would require unrealistically long model run times, and may cause numerical problems.
3.3 Results of FLAC3D Analyses

Results from the FLAC3D analyses were compared to (1) the published results of the field
measurements, and (2) the published results of the two-dimensional linear elastic analyses
performed by Kolisoja et al. (2000) of their own field tests.
The vertical pressure with depth beneath a train with an axle load of 250 kN is shown in Figure
3.4. Instrumentation data was provided at discrete points beneath the train load; data of the
longitudinal distribution at each depth is not available. As can be seen in Figure 3.4, both the
field measurements and numerical analyses indicate that the additional vertical pressures
caused by the passing of a train decrease with depth. The greatest increase in vertical stress
is concentrated within the upper meter of the embankment. Kolisoja and Mkel (2001) point
out that one pressure cell appears to give an unreliable measurement; however, in general,
the authors state that the instrumentation performed well and that meaningful results were
obtained even from layers that, at the beginning, were considered to be too coarse grained for
any instrumentation.
Vertical Pressure (kPa)
0 40 80 120 160 200

0
-0.5
Depth from the Bottom of Sleeper (m)
-1
-1.5
-2 Pressure cell data

Calculated by Kolisoja and Mkel (2001)
FLAC3D data
-2.5
Figure 3.4. Vertical pressure with depth for a train axle load of 250 kN
The pressure cell measurements at a depth of 0.5 m below the bottom of the sleepers for vari-
ous axle loads are shown in Figure 3.5. The data in Figure 3.5 include measurements taken
beneath four crossings of the test train with an axle load of 250 kN, various tank trains, and
one crossing of maintenance equipment with four exceptionally light axle loads (Kolisoja and
Mkel 2001). As can be seen in Figure 3.5, there is a relatively linear trend between the
measured vertical pressure and axle load, though there is some scatter in the data. The re-
sults of the two-dimensional numerical analyses by Kolisoja et al. (2000) are shown on Figure
3.5 as the black circles. Data was also presented in Finnish by Kolisoja et al. (2000) for pres-
sure cell measurements taken at a depth of 1.0 m below the bottom of the sleepers.
Figure 3.5. Vertical pressures at a depth of 0.5 m below the sleeper

as a function of axle load
(from Kolisoja and Mkel 2001).
100
80
Kolisoja and Mkel (2001), depth 0.5 m

60 FLAC3D, depth 0.5 m
Kolisoja and Mkel (2001), depth 1.0 m
FLAC3D, depth 1.0 m
40
20
0
50 100 150 200 250 300 350
Axle Load (kN)
Figure 3.6. Vertical pressure at depths of 0.5 and 1.0 m as a function of axle load
The vertical pressures calculated by FLAC3D at depths of 0.5 and 1.0 m below the bottom of
the sleeper for various axle loads are presented in Figure 3.6. The solid lines in Figure 3.6
are the results of the analyses performed by Kolisoja et al. (2000), which also seem to serve
as "best fit" lines for their field measurements (see Figure 3.5). As can be seen in Figure 3.6,
there exists a fairly linear relationship between vertical pressure and train axle load for the
range of axle loads investigated. Also, it can be seen here that vertical pressures decrease
with depth. Furthermore, as the train load increases, the difference between the stresses at
a depth of 0.5 m and 1.0 m increase. It is reasonable to assume that for the range of axle
loads investigated, the greatest increase in vertical stresses remain concentrated in the up-
per meter of embankment material.
As can be seen in Figures 3.4 and 3.6, the FLAC3D analyses provide a good fit to both the
field pressure cell measurements and the analyses performed by Kolisoja et al. (2000).
3.4 Closing Remarks

The results of FLAC3D dynamic analyses were compared with data from pressure cells in-
stalled at a test embankment along the railway line between Koria and Kouvola, Finland.
Both the field measurements and FLAC3D analyses indicated that:
(1) the vertical stresses beneath a train decrease with depth, with most of the increase in
vertical stress concentrated within a depth of 1.0 m below the bottom of the sleeper,
and
(2) vertical pressures within the embankment system increase with increasing train axle
load.
The FLAC3D analyses provided a good fit to the pressure cell data. This verification study
demonstrated that:
(1) Reasonable results are achieved when a single bogie is considered, and the train
load is input dynamically as a cosine wave, and
(2) Though in reality the sleepers distribute the train load to the embankment system,
reasonable results were achieved in the numerical analyses despite the fact that no
structural elements were used to represent the rail or sleepers.
At low train velocities, the ground response from a moving load is essentially quasi-static;
however, stresses and vertical movements increase substantially as the train velocity ap-
proaches the Rayleigh-wave velocity of the soil (Adolfsson et al. 1999b, Kaynia et al. 2000,
Chen et al. 2005). The field measurements described by Kolisoja and Mkel (2001) in-
volved the use of trains with velocities ranging from 40 to 100 km/hr. At these velocities, the
Rayleigh-wave velocity of the embankment system was not surpassed; therefore, the system
responded in a quasi-static manner. As shown in the following section, as train velocities
approach the Rayleigh-wave velocity of the embankment system, deformations and stresses
increase non-linearly.
4 NUMERICAL ANALYSES WITH CONCRETE BRIDGE
The FLAC3D analyses described in the previous section provided verification and an under-
standing of the appropriate numerical modelling procedures for a railway embankment system
subjected to moving train loads. The same numerical modelling procedures were adopted for
the case in which a bridge exists along the railway. The aim of the analyses is to evaluate the
stress distribution and deflections within the transition zone of a railway bridge due to the
passing of train with high speeds and high axle loads.
A numerical parametric study was performed to study the effects of 1.) the train load, 2.) the
train velocity, and 3.) the stiffness of the ballast, subballast and embankment fill. The ap-
proach adopted for the numerical parameter studies relies on a base case analysis, with sys-
tematic variation of parameter values from the base case. The parametric study was per-
formed using a typical concrete bridge geometry. This section provides a description of the
results of the numerical parameter study, including a discussion of the trends disclosed by the
analyses.
4.1 FLAC3D Model and Analyses

Numerical analyses of a simplified concrete bridge geometry were performed using the dy-
namic FLAC3D computer program. The bridge geometry is based upon a typical concrete
bridge geometry as provided by the Swedish Railroad Administration, Banverket. The mesh
geometry of the concrete bridge model is shown in Figure 4.1. The total length of the model
is 56 m, and the length of the bridge is 10 m.
Figure 4.1. Mesh geometry of concrete bridge numerical model.

The embankment profile of the bridge geometry is shown in Figure 4.2. The profile consists
of an upper ballast layer 0.4 m thick, which is underlain by a layer of subballast 1.2 m thick,
which in turn is underlain by embankment backfill. A stiff, non-yielding, bearing layer is as-
sumed to exist beneath the embankment fill. As can be seen in Figure 4.2, the profile is
symmetrical about its centre-line.
7.4 m
3.3 m Train Load
2.5 m
Ballast
0.4 m
Subballast 2 1.2 m
1
Embankment Fill
3.8 m
Stiff Bearing Layer

27.4 m
Figure 4.2. Cross-section of numerical models.
elastic perfectly plastic model with a Mohr-Coulomb failure criterion was used to represent the
ballast, subballast, and embankment fill.
Table 4.1. Material property values used in the numerical analyses.
Ballast Subballast Fill

Dry Density (kg/m3) 1900 1900 1700
Elastic Modulus (MPa) 193 160 47.9
Poissons Ratio 0.30 0.30 0.31
Bulk Modulus (MPa) 161 133 42
Shear Modulus (MPa) 74 61.5 18.3
Friction Angle, (deg) 40 37 31
Dilation Angle (deg) 4 3 2
No structural elements were used to model the bridge abutments, or the rail and sleepers. It
was desired to model the train dynamically, and since the timestep in the dynamic mode is
determined by the largest material stiffness, using structural elements would require unrealis-
tically long model run times, and may cause numerical problems. Therefore, the bridge struc-
ture was represented by fixed gridpoints at the bridge-soil interface. Since deflections at the
structure beneath the passing train are on the order of a few millimetres, these boundary
conditions are not expected to significantly affect the results.
The train load was applied dynamically along the top of the model over a width of 1.25 m from
the centre-line. To gain a clear understanding of the behaviour of the system under a dynamic
load, only one bogie of the train was considered. The simplified approximation of the load be-
neath two wheels of one bogie is similar to that shown in Figure 3.3, depending upon axle
spacing.
Results are presented at the first bridge abutment when the train is directly above the back of
the abutment (see Figure 4.3). The first abutment is the end of the bridge first encountered
by the train as the train moves toward and over the bridge. Results are presented for the first
abutment because there are significant increases in stresses exerted on the bridge, and de-
flections behind the bridge, that occur as the train approaches the bridge. This is due to the
fact that the train is moving from the railway embankment onto the stiff bridge structure. The
bridge structure is less affected by the moving train as the train moves from the bridge struc-
ture onto the railway embankment.
Position of bogie
for which results
are reported
Figure 4.3. Location of first bridge abutment in model
4.2 Results of Parametric Analyses

A numerical parametric study was performed using the FLAC3D model to study the effects of
1.) the train load, 2.) the train velocity, and 3.) the stiffness of the ballast, subballast and em-
bankment fill. The results of the parametric study are discussed in the following sections.
4.2.1 Train Axle Load

For the parameter study, four train loads were considered:
(1) Passenger train, representative of the Swedish X2000 high speed train, with an axle
load of 125 kN and a wheel spacing of 2.9 m,
(2) Passenger train with an axle load of 250 kN and a wheel spacing of 2.9 m,
(3) Freight train with an axle load of 225 kN and a wheel spacing of 1.8 m, and
(4) Freight train with an axle load of 330 kN and a wheel spacing of 1.8 m.
Typically, the axle loads of freight trains are greater and the wheel spacings are less than
those of passenger trains. The passenger train loads were applied over a range of velocities
from 50 to 350 km/hr, and the freight trains were applied at velocities of 50 and 100 km/hr. A
velocity of 100 km/hr is considered to be an upper limit velocity for a freight train.
(a) Paxle = 125 kN
(b) Paxle = 250 kN

Figure 4.4. Distribution of net horizontal stresses behind first bridge abutment
(v = 350 km/hr) (Contours given in kPa)
The increase in horizontal stresses in the embankment fill material occurs slightly ahead of
the moving train. However, the greatest increases in horizontal and vertical stresses occur in
the ballast and subballast layers; these increases occur approximately beneath the train
when the train is above the back of the abutment. A cross-section of the net horizontal
stresses on the first bridge abutment for axle loads of 125 and 250 kN are shown in Figure
4.4 for the case in the train velocity was equal to 350 km/hr.
As can be seen in Figure 4.4, the maximum net horizontal stresses occur between the depths
of 0.2 to 0.45 m. For all the analyses in this parametric study, the net maximum horizontal
stresses were observed within this range of depth, which means that in all cases, the maxi-
mum net horizontal stresses occur in the ballast layer or at the ballast-subballast layer inter-
face. For both cases, below a depth of about 1 to 1.2 m, the net horizontal stresses in the
embankment fill are relatively small. However, since the increase in horizontal stresses in the
embankment fill material occurs slightly ahead of the moving train, horizontal stresses at this
depth are slightly greater as the train is approaching the abutment.
The FLAC3D analyses indicate that as the train axle load increases, the vertical and horizontal
stresses within the railway system increase. Figure 4.5 shows the vertical pressures behind
the first concrete bridge abutment calculated by FLAC3D at depths of 0.2 and 0.4 m for vari-
ous axle loads traveling at a velocity of 50 km/hr. Figure 4.6 shows the horizontal pressures
calculated at depths of 0.2 and 0.4 m for (1) various axle loads traveling at a velocity of 50
km/hr and (2) axle loads of 125 and 250 kN traveling at 350 km/hr. It should be noted that the
ballast layer is 0.4 m thick, and the top of the bridge structure is located at the base of the bal-
last layer.
As can be seen in Figure 4.5, the vertical pressures are greater at the shallower depth of 0.2
m, which is expected. In Figure 4.6, it can be seen that there is not much difference in the
horizontal pressures between the depths of 0.2 and 0.4 m for a given train velocity; however,
horizontal pressures increase as train velocity increases. The effects of train velocity on verti-
cal and horizontal stresses are discussed in greater detail in the following section.
100 z = -0.2 m
z = -0.4 m
80
60
40
20
0
100 150 200 250 300 350
Axle Load (kN)
Figure 4.5. Vertical pressure at depths of 0.2 and 0.4 m (v = 50 km/hr)

80
Horizontal Pressure (kPa)

60
40
v = 50 km/hr, z = -0.2 m
20
v = 50 km/hr, z = -0.4 m
v = 350 km/hr, z = -0.2 m
v = 350 km/hr, z = -0.4 m
0
100 150 200 250 300 350
Axle Load (kN)
Figure 4.6. Horizontal pressure at depths of 0.2 and 0.4 m
A cross-section of the vertical deflections 1 m behind the first bridge abutment is shown in Fig-
ure 4.7 for the case in which the train axle load was equal to 250 kN, the train velocity was
equal to 350 km/hr, and the base case material properties were used. When the train is di-
rectly above the back of the abutment, the maximum vertical deflections occur approximately
0.6 to 1 m behind the train (unless noted otherwise). For the case shown in Figure 4.7, the
maximum vertical deflection behind the first bridge abutment is 2.87 mm.
A few general observations can be made based on the results shown in Figure 4.7. First,
most of the vertical deflections occur in the ballast and subballast layers. Second, as would be
expected, the maximum vertical deflections are concentrated beneath the width of the train.
As the train axle load increases, the maximum vertical deflections increase. The vertical de-
flections for axle loads of 125 and 250 kN for a wide range of train velocities is discussed in
the next section.
Figure 4.7. Distribution of vertical deflections 1 m behind first bridge abutment

for base case conditions (Paxle = 250 kN, v = 350 km/hr)
4.2.2 Train Velocity

The passenger train loads were applied over a range of velocities from 50 to 350 km/hr, and
the freight trains were applied at velocities of 50 and 100 km/hr. The passenger and freight
trains have axle spacings of 2.9 and 1.8 m, respectively. The base case material property val-
ues, listed in Table 4.1, were not varied. A plot of the train velocity versus maximum net hori-
zontal stress is shown in Figure 4.8, and plot of the train velocity versus maximum vertical de-
flection behind the first abutment is shown in Figure 4.9.
The values shown in Figures 4.8 and 4.9 correspond to when the train is located directly
above the back of the first abutment. Results are expressed in terms of net horizontal stress,
which represents the additional horizontal stress applied to the back of the structure due to the
train load, and maximum vertical deflections behind the abutment. The maximum net horizon-
tal stresses occur between depths of 0.2 and 0.45 m, and the maximum vertical deflections
occur approximately 0.6 m behind the train.
80
Maximum Net Horizontal Stress (kPa) 60
Paxle = 125 kN
Paxle = 225 kN (freight)
40
Paxle = 250 kN
20
0
0 100 200 300 400
Train Velocity (km/hr)
Figure 4.8. Train velocity versus maximum net horizontal stress
3
Maximum Vertical Deflection (mm)
2
Paxle = 125 kN
Paxle = 250 kN
1
0
0 100 200 300 400
Figure 4.9. Train velocity versus maximum vertical deflection
In general, as the train velocity increases, the net horizontal stresses in the ballast and subbal-
last layer increase. However, the relationship between train velocity and maximum net hori-
zontal stress on the back of the abutment is not linear. For the passenger trains, the largest
incremental increase in the maximum net horizontal stresses at the first bridge abutment oc-
curs when the train velocity increases from 250 to 350 km/hr.
Similarly, as the train velocity increases, the maximum vertical deflection behind the train in-
creases. The relationship between train velocity and maximum vertical deflection behind the
first abutment is also not linear; the largest increase in maximum vertical deflections occurs for
the case in which the train velocity increases from 250 to 350 km/hr.
As can be seen in Figures 4.8 and 4.9, the non-linear relationships between train velocity
and maximum net horizontal stress and vertical deflections are more pronounced for the
heavier train axle load.
The distribution with depth of vertical and horizontal pressures caused by the passage of a
passenger train with an axle load of 250 kN at velocities of 50 and 350 km/hr are shown in
Figures 4.10 and 4.11, respectively. The following observations can be made based upon the
results shown in Figures 4.10 and 4.11:
Within the upper meter, both the vertical and horizontal stresses increase with increas-
ing train velocity; though the increase is relatively small for the vertical pressures. This
would indicate that the vertical pressures are affected more by train axle load, and the
horizontal pressures by train velocity. These trends can also be seen in Figure 4.5. 4.6.
As train velocity increases, the horizontal stresses increase. This can also be seen in
Figure 4.6 as a function of train axle load.
Below a depth of about 1.6 m, which is the bottom of the subballast layer, there is little
variation in horizontal stresses for both train velocities;
For a train velocity of 350 km/hr, negative net vertical and horizontal pressures are ob-
served around a depth of about 1.5 m. Negative values of pressure indicate that the
soil is in tension. The values of tension, however, are relatively small.

0 20 40 60 80 100
0
-1
-2
Depth (m)
-3
at bridge, v = 50 km/hr
at bridge, v = 350 km/hr
-4
-5
Figure 4.10. Variation in Vertical Pressure with Depth

Horizontal Pressure (kPa)

0 20 40 60 80 100
0
-0.4
-0.8
Depth (m)
-1.2
v = 50 km/hr
v = 350 km/hr
-1.6
-2
Figure 4.11. Variation in Horizontal Pressure with Depth
4.2.3 Material Property Values of Ballast, Subballast, and Embankment Fill

Analyses were performed to evaluate the effect of the material property values on the load
distribution and vertical deflections within the transition zone. For the parametric study, the
stiffness, or Elastic Modulus, of the ballast, subballast, and embankment fill were varied sys-
tematically according to the values listed in Table 4.2.
Table 4.2. Variable material property values

Rayleigh Wave
E (MPa) Poissons Ratio
Velocity (km/hr)
160 0.30 600
Ballast 193 0.30 660
300 0.30 820
47.9 0.31 345
Subballast 160 0.30 600
300 0.30 820
20 0.31 225
Embankment
47.9 0.31 345
Fill
70 0.30 420
(Bold values are base case values.)
The Rayleigh wave velocity for each material is also provided in Table 4.2. The Rayleigh
wave velocity, VR, of a material may be estimated using the following expression (Hall 2000):
G (0.87 + 1.12 )
VR =
(1 + )
E
Where G = shear modulus = , E = Elastic Modulus, = Poisson's ratio, and =
2 (1 + )
density.
The goal of this research is to evaluate the behaviour of bridge systems under high speed
trains, which may have velocities up to 350 km/hr in the future. The parametric analyses
described in the following sections were performed using a train velocity of 350 km/hr, and
axle loads of 125 and 250 kN. The variation in the maximum net horizontal stresses, net
h,max, with velocity and material stiffness values is given in Table 4.3. The values of maxi-
mum net horizontal stresses in Table 4.3 were observed between depths of 0.2 and 0.45 m,
and are for the case in which the train is directly above the back of the first abutment. The
values in Table 4.3 are averaged values over a distance of 1 m from the center line of the
abutment. The variation of net horizontal stresses with material property values is discussed
in the following sections.
The variation in the maximum vertical deflections, v,max, with train velocity and material stiff-
ness values is given in Table 4.4 for the case in which the train is above the back of the
abutment, and is discussed in the following sections.
Table 4.3. Maximum net horizontal stresses at the first abutment
Paxle = 125 kN Paxle = 250 kN

% change % change
Net h,max Net h,max
Case from base from base
(kPa) (kPa)
case case
v = 50 km/hr 13.7 31.5
v = 150 km/hr 14.3 34.9
v = 250 km/hr 19.3 43.5
v = 350 km/hr 31.5 72.7
Ballast (E=160 MPa) 31.8 negligible 78.4 +7.8
Ballast (E=300 MPa) 31.3 negligible 65.3 -10.2
Subballast (E=47.9 MPa) 43.0 +36.5 87.5 +20.4
Subballast (E=300 MPa) 24.1 -23.5 40.1 -44.8
Fill (E=20 MPa) 36.1 +14.6 76.5 +5.2
Fill (E=70 MPa) 17.4 -44.8 41.8 -42.5
1
Bold value is base case value.
Table 4.4. Maximum vertical deflections at the first abutment
Paxle = 125 kN Paxle = 250 kN

% change % change
v,max v,max
Case from base from base
(mm) (mm)
case case
v = 50 km/hr 0.50 1.45
v = 150 km/hr 0.52 1.49
v = 250 km/hr 0.62 1.93
v = 350 km/hr 0.83 1 2.87 1
Ballast (E=160 MPa) 0.90 +8.4 3.03 +5.6
Ballast (E=300 MPa) 0.73 -12.0 2.71 -5.6
Subballast (E=47.9 MPa) 1.89 +127.7 5.46 +90.2
Subballast (E=300 MPa) 0.46 -44.6 1.83 -36.2
Fill (E=20 MPa) 3.05 +267.5 2 9.10 +217 2
Fill (E=70 MPa) 0.48 -42.2 1.43 -50.1
1
Bold value is base case value.
2
Observed 3 m behind train.
Ballast stiffness
The elastic modulus of the ballast was varied over a range of 160 to 300 MPa. For an axle
load of 125 kN, changing the stiffness of the ballast material has a relatively small effect on
the net horizontal stresses behind the first bridge abutment (see Table 4.3). For an axle load
of 250 kN, as the stiffness of the ballast increases, the net horizontal stresses decrease. For
both axle loads, an increase in ballast stiffness leads to a decrease in maximum vertical de-
flections behind the bridge abutment.
Subballast Stiffness
The elastic modulus of the subballast was varied over a range of 47.9 to 300 MPa. When
the stiffness of the subballast layer is equal to 47.9 MPa, it is equal to the stiffness of the
embankment fill, and in this case, it is as if there is no subballast layer and the embankment
fill extends to the bottom of the ballast layer. When the stiffness of the subballast layer is
equal to 300 MPa, the subballast is 1.6 times stiffer than the ballast layer.
The stiffness of the subballast layer has a significant effect on the calculated maximum net
horizontal stresses, which occur at the ballast-subballast interface. For both axle loads, the
greatest increase in net horizontal stresses occurs for the case in which the softer subballast is
used. For this case, the Rayleigh wave velocity of the subballast and embankment fill is on
the order of 345 km/hr. Conversely, for both axle loads, as the stiffness of the subballast in-
creases, the maximum horizontal stresses decrease. Furthermore, for an axle load of 250
km/hr the decrease in horizontal stress due to a stiffer subballast layer is greater than the de-
crease in stress that is associated with decreasing the train velocity from 350 to 250 km/hr.
A cross-section of the net horizontal stresses at the first bridge abutment are shown in Figure
4.12 for the case in which the stiffest (E = 300 MPa) subballast layer was used and the train
velocity was equal to 350 km/hr. As can be seen in Figure 4.12, the maximum net horizontal
stresses occur at a depth of about 0.4 m, at the ballast-subballast layer interface. The net
horizontal stresses below a depth of 1 m are relatively small.
(a) Paxle = 125 kN
(b) Paxle = 250 kN
Figure 4.12. Net horizontal stresses behind first bridge abutment (v=350 km/hr)
for stiff subballast case. (Contours given in kPa.)
As can be seen in Table 4.4, as the stiffness of the subballast layer increases, vertical deflec-
tions behind the first abutment decrease.
Embankment Fill
The elastic modulus of the embankment fill was varied over a range of 20 to 70 MPa. For
both axle loads, an increase in the stiffness of the fill corresponds to a decrease in the net
horizontal stresses in the ballast and subballast layers. Furthermore, for both axle loads, the
decrease in horizontal stress due to a stiffer embankment fill layer is greater than the de-
crease in stress that is associated with decreasing the train velocity from 350 to 250 km/hr.
A cross-section of the net horizontal stresses at the first bridge abutment are shown in Figure
3.13 for the case in which the softest fill layer was used and the train velocity was equal to
350 km/hr. For the heavier axle load, a decrease in stiffness of the embankment fill has little
effect on the calculated magnitude of maximum net horizontal stress. As can be seen in Fig-
ure 4.13, the maximum net horizontal stress occurs at about a depth of 0.2 m and is ap-
proximately 76 kPa. However, net horizontal stresses on the order of 10 kPa are observed to
a depth of approximately 3.8 m. The maximum increase in stresses in the embankment fill,
however, occurs slightly ahead of the train. Net horizontal stresses on the order of 30 kPa
were observed at the subballast-fill interface slightly ahead of the moving train for this case.
For both axle loads, the greatest vertical deflection behind the first bridge abutment was cal-
culated for the case in which the soft embankment fill was used. This high value of calculated
deflection occurs approximately 3 m behind the train.
(a) Paxle = 125 kN
(b) Paxle = 250 kN
Figure 4.13. Net horizontal stresses behind first bridge abutment (v=350 km/hr)
for soft fill case. (Contours given in kPa.)
4.3 Summary and Conclusions

Numerical analyses were performed using FLAC3D to evaluate the behaviour of bridge tran-
sition zones under the passing of trains. A bridge geometry representative of a typical Euro-
pean concrete bridge was considered. Results were given for the vertical and horizontal
stresses on the back of the first bridge abutment, and the maximum vertical deflections at the
first bridge abutment. A numerical parametric study was performed to study the effects of 1.)
the train load, 2.) the train velocity, and 3.) the stiffness of the ballast, subballast and em-
bankment fill.
The greatest increase in vertical and horizontal stresses occur in the ballast and sub-
ballast layers; these increases occur approximately beneath the train when the train is
above the back of the abutment.
upper 1.0 to 1.2 m of the system. The vertical and horizontal stresses caused by the
passing of a train decrease with depth. Similar observations have been made in field
tests by Kolisoja et al. (2000).
tions beneath the train increase. Similar observations have been made by Katzenbach
and Ittershagen (2005) and Hunt (1997), respectively.
Within the upper meter, both the vertical and horizontal stresses increase with increas-
ing train velocity; though the increase in vertical pressure is relatively small. The verti-
cal pressures are affected more by train axle load, and the horizontal pressures by train
velocity. Additionally, there is a substantial increase in both the maximum horizontal
stresses and the vertical deflections when the train velocity increases from 250 to 350
km/hr.
stresses exerted on the bridge abutment decrease. Horizontal stresses in the em-
bankment fill are relatively small, but increase when a soft fill is used.
Changing the stiffness of upper ballast material has little effect on the calculated verti-
cal deflections behind the first bridge abutment. This is probably due to the fact that
the ballast layer is relatively thin.
As the stiffness of the subballast layer increases, vertical deflections behind the first
abutment decrease. For the majority of the cases evaluated, when the train is located
directly above the back of the first bridge abutment, the maximum vertical deflections
are observed to occur approximately 1 m behind the train, except for the case in which
a soft fill is used. When a soft embankment fill is used, vertical deflections increase
substantially, and these deflections are observed to occur 3 m behind the train, as if
there is a stern wave behind the train. The presence of a wave-like phenomenon in
front of and in back of the train has also been documented by Adolfsson et al. (1999b)
and Kopf and Adam (2005).
5 NUMERICAL ANALYSES WITH MASONRY ARCH BRIDGE
5.1 FLAC3D Model and Analyses

A series of numerical analyses were also performed using a typical masonry arch bridge ge-
ometry. The model geometry is based upon a typical masonry arch bridge geometry as pro-
vided by Salford University, and is shown in Figure 5.1. The total length of the model is 54
m; the arch is assumed to sit on abutments that are 3 m tall, has a radius of 4.9 m and a total
span of 8.4 m. Taking into account a masonry arch ring thickness of 70 cm, the span-to-rise
ratio of the arch is approximately 4.
Figure 5.1. Mesh geometry of numerical model of masonry arch bridge
Like the concrete bridge geometry, the masonry arch bridge geometry has the embankment
profile that is shown in Figure 5.2. The profile consists of an upper ballast layer 0.4 m thick,
which is underlain by a layer of subballast 1.2 m thick, which in turn is underlain by embank-
ment backfill. A stiff, non-yielding, bearing layer is assumed to exist beneath the embank-
ment fill. As can be seen in Figure 5.2, the profile is symmetrical about its centre-line.
7.4 m
3.3 m Train Load
2.5 m
Ballast
0.4 m
Subballast 2 1.2 m
1
Embankment Fill
3.8 m
Stiff Bearing Layer

27.4 m
Figure 5.2. Cross-section of numerical models.

elastic perfectly plastic model with a Mohr-Coulomb failure criterion was used to represent the
ballast, subballast, and embankment fill.
Table 5.1. Material property values used in the numerical analyses.
Ballast Subballast Fill

Dry Density (kg/m3) 1900 1900 1700
Elastic Modulus (MPa) 193 160 47.9
Poissons Ratio 0.30 0.30 0.31
Bulk Modulus (MPa) 161 133 42
Shear Modulus (MPa) 74 61.5 18.3
Friction Angle, (deg) 40 37 31
Dilation Angle (deg) 4 3 2
No structural elements were used to model the bridge, or the rail and sleepers. It was desired
to model the train dynamically, and since the timestep in the dynamic mode is determined by
the largest material stiffness, using structural elements would require unrealistically long
model run times, and may cause numerical problems. Therefore, the bridge structure was
represented by fixed gridpoints at the bridge-soil interface.
A series of numerical analyses were also performed using the masonry arch bridge geometry
shown in Figure 5.1. Three train load and velocity configurations were considered:
1. The high speed X2000 train with an axle load of 125 kN, axle spacing of 2.9 m, and
velocity of 350 km/hr,
2. A freight train with an axle load of 225 kN, axle spacing of 1.8 m, and velocity of 100
km/hr, and
3. A freight train with an axle load of 330 kN, axle spacing of 1.8 m, and velocity of 100
km/hr.
5.2 Results of Analyses

Cross-sections of net vertical and horizontal stresses in the material above the arch are
shown for the three train configurations in Figures 5.4 through 5.6. The results shown in Fig-
ures 5.4 through 5.6 are taken at the first quarter section of the arch, the location of which is
Figure 5.3. Location of cross-sections shown in Figures 5.4 through 5.6.
(a) Net Vertical Stresses
(b) Net Horizontal Stresses
Figure 5.4. Net stresses in arch bridge under passing of high speed X2000 train,
Paxle = 125 kN, v = 350 km/hr (contours in kPa)

Figure 5.5. Net stresses in arch bridge under freight train

Figure 5.6. Net stresses in arch bridge under freight train
350 350
300 300
250 250

200 200
150 150
100 100
50 50
150 200 250 300 150 200 250 300
Axle Load (kN) Axle Load (kN)
(a) Net Horizontal Stresses (b) Net Vertical Stresses
Figure 5.7. Arch bridge geometry: projected contours of net stress at a depth of 0.5 m
as a function of train velocity and axle load
Projected contours of net horizontal and net vertical stress at the point of the arch at a
depth of 0.5 were developed. Plots of the contours are shown in Figure 5.7. As can be seen
in Figure 5.7, at a depth of about 0.5 m, net horizontal stresses increase with increasing train
velocity and axle load; however increasing the train velocity over a range of about 50 to 150
km/hr has little affect on the horizontal stresses for a given axle load. At a depth of 0.5 m,
the vertical stresses increase with axle load and are not affected significantly by increases in
train velocity.
4
Normalized Vertical Stress
6
10
12
at crown
14 at 3/4 point
16
1.6 1.8 2 2.2 2.4 2.6 2.8 3

Time (s)
(a) Axle Load = 225 kN, velocity = 50 km/hr
4
10
12
at crown
14 at 3/4 point
16
0.9 1 1.1 1.2 1.3

Time (s)
(b) Axle Load = 225 kN, velocity = 100 km/hr
Figure 5.8. Influence lines at the crown and point of arch bridge (Axle Load = 225 kN)
4
6
8
10
12
14
16
18
20 at crown
at 3/4 point
22
24
1.8 2 2.2 2.4 2.6

Time (s)
(a) Axle Load = 330 kN, velocity = 50 km/hr
4
6
10
12
14
16
18
20 at crown
at 3/4 point
22
24
0.9 1 1.1 1.2 1.3

Time (s)
(b) Axle Load = 330 kN, velocity = 100 km/hr
Figure 5.9. Influence lines at the crown and point of arch bridge (Axle Load = 330 kN)
Influence lines, in the form of normalized vertical stress against time, are shown in Figures
5.8 and 5.9 for trains with axle loads of 225 and 330 kN, respectively. The influence lines
shown in Figures 5.8 and 5.9 pertain to points directly above the crown and the point of
the arch. As an example, under the passing of a train with an axle load of 225 kN and a ve-
locity of 100 km/hr, the arch structure will experience a relative increase in vertical stress at
the crown on the order of about 16 times the initial stress.
The normalized vertical stress in Figures 5.8 and 5.9 is defined as the vertical stress induced
by the passing train divided by the initial (static) vertical stress. It should be noted that the
initial vertical stresses at the and points are higher than that at the crown. Therefore,
even though the crown feels a greater relative increase in vertical stress, the total vertical
stresses induced at the crown and the and points are similar in total magnitude. As has
been demonstrated previously, the induced vertical stresses decrease with depth; in most
cases, most of the induced vertical stresses due to a passing train are concentrated within
the upper 1.0 to 1.5 m.
There are a few general observations that can be made based upon Figures 5.8 and 5.9.
First, as would be expected, the relative vertical stresses induced on the arch structure in-
crease with train axle load. However, vertical stresses are not greatly affected by train veloc-
ity. Second, since there is less soil above the crown of the arch to dissipate stresses, the
crown of the arch will experience the greatest spike in vertical stress under the passing of a
train.
5.3 Summary and Conclusions

It can be assumed that the factors that affect the stress distribution and deflections for the
concrete bridge structure (as described in Section 4.0) will have the same effect for the ma-
sonry arch bridge structure, such as:
The greatest increase in vertical and horizontal stresses occur in the ballast and sub-
ballast layers; these increases occur approximately beneath the train when the train is
above the bridge.
Horizontal stresses increase with increasing train velocity and axle load. The vertical
pressures are not significantly affected by train velocity for the range of velocities
evaluated. However, vertical pressures may be expected to increase at very high train
velocities (i.e. train velocities that approach the critical velocity of the bridge and em-
bankment system.)
The geometry of the arch bridge is different than the concrete bridge in two significant ways:
(1) for masonry arch bridges there is a gradual change in stiffness, while for a concrete
bridge the change is more abrupt, and (2) for the masonry arch structure, the ballast and
subballast are confined within a relatively small volume. Due to the latter factor, additional
vertical and horizontal stresses induced by the train are distributed immediately to the struc-
ture within the arch. A few additional general observations can be made based upon the
analyses using the masonry arch bridge:
As the train axle load increases, the vertical and horizontal stresses below the train in-
crease, as expected. Significant vertical stresses are exerted on the top of the arch.
These stresses are concentrated primarily beneath the width of the train.
For a passenger train with an axle load of 125 kN, the maximum net horizontal stresses
occur to depths of about 0.7 m; however, for the freight trains with axle loads of 225
and 330 kN, net horizontal stresses on the order of 15 and 20 kPa, respectively, are
observed down to depths of 0.9 to 1.0 m.
6 LIMITATIONS
tory tests were performed. This document is a background document, intended to provide
insight into the general behavior of bridge systems under the passing of high speed trains.
No design methods or recommendations are presented herein.
The results presented in this document are only for the conditions investigated. The findings
of the numerical analyses should be verified against laboratory scaled tests and/or full-scale
field tests.
7 CONCLUSIONS
tory tests were performed. Numerical analyses consisted of three-dimensional dynamic
analyses. Using a bridge geometry representative of a typical European concrete bridge, a
numerical parametric study was performed to study the effects of 1.) the train load, 2.) the
train velocity, and 3.) the stiffness of the ballast, subballast and embankment fill. Numerical
analyses were also performed using a bridge geometry representative of European masonry
arch bridges.
upper 1.0 to 1.2 m of the system. The additional vertical and horizontal stresses
caused by the passing of a train decrease with depth.
As train velocity increases, the horizontal stresses and vertical deflections increase,
however, they do not increase linearly with increasing velocity. As the velocity of the
train approaches the Rayleigh wave velocity of the embankment system, significant
horizontal stresses and vertical deflections develop.
stresses exerted on the bridge decrease. Horizontal stresses in the embankment fill
are relatively small, but increase when a soft fill is used.
When a soft embankment fill is used, vertical deflections increase substantially, and
these deflections are observed to occur 3 m behind the train, as if there is a stern
wave behind the train.
The numerical methods used in this research were verified against data from an instru-
mented test embankment published by the Finnish railway administration. However, the Fin-
nish test embankment did not include a bridge structure. The results presented in this report
are intended to provide insight into the behavior of railway bridge transition zones. The re-
sults presented in this report are valid for the conditions studied; however, field tests and/or
laboratory tests are needed to verify the numerical analyses.
8 REFERENCES
Adolfsson, K., Andrasson, B., Bengtsson, P., and Zackrisson, P. (1999a). High speed train
X2000 on soft organic clay measurements in Sweden. Proceedings of the European
Conference on Soil Mechanics and Geotechnical Engineering, Geotechnical engineering
for transportation infrastructure, Amsterdam, June 1999. Vol. 3: 1713 1718.
Adolfsson, K., Andrasson, B., Bengtsson, P-E., Bodare, A., Madshus, C., Massarsch, R.,
Wallmark, G., & Zackrisson, P. (1999b). Evaluation and Analyses of Measurements from
the West Coast Line. Swedish Geotechnical Institute Report.
Andrasson, B. (2002). Development of Tools Track and Environmental Vibrations, Use of

the FLAC and FLAC3D Programs. Joint Nordic Railway Vibration Research Project,
NORDVIB, J&W Report, Gteborg, Sweden.
Chen, Y.M., Wang, C.J., Chen, Y.P., and Zhu, B. (2005). Characteristics of stresses and set-
tlement of ground induced by train. Environmental Vibrations, Takemiya (ed.) Taylor and
Francis Group, London: 33 42.
Cojocaru, E.C., Irschik, H., and Schlacher, K. (2003). Concentrations of Pressure Between
an Elastically Supported Beam and a Moving Timoshenko-Beam. Journal of Engineering
Mechanics, 129, no. 9, 1076-1082.
ERRI (European Rail Research Institute) (1999). State of the Art Report, Bridge Ends, Em-
bankment Structure Transition. ERRI Report D 230.1/RP3, Utrecht, 100 pp.
Hall, L. and Bodare, A. (1999). Prediction of frequency content in train induced ground vibra-
tions based on a beam-on-Winkler-foundation model. Proceedings, Geotechnical Engi-
neering for Transportation Infrastructure, Barends et al. (eds), Balkema, Rotterdam: 1803
- 1808.
Hall (2000). Simulations and Analyses of Train-Induced Ground Vibrations. Doctoral Thesis
1034, Division of Soil and Rock Mechanics, Department of Civil and Environmental Engi-
neering, Royal Institute of Technology, Stockholm.
Hunt, H.E.M. (1997). Settlement of railway track near bridge abutments. Proceedings Inst.
Civil Engs., Transp., 123, no. 1: 68-73.
Itasca (2005). FLAC3D Fast Lagrangian Analysis of Continua in 3 Dimensions, Users Guide,
Itasca Consulting Group, Inc., Minneapolis, Minnesota, USA.
Katzenbach, R. And Ittershagen, M. (2005). New developments in soil improvement under

railway lines on soft clay. Proceedings of the 16th International Conference on Soil Me-
chanics and Geotechnical Engineering, Osaka, September, 2005. Vol. 3: 1207 1210.
Kaynia, A.M., Madshus, C., and Zackrisson, P. (2000). Ground Vibrations From High-Speed
Trains: Prediction and Countermeasure. Journal of Geotechnical and Geoenvironmental
Engineering, 126, no. 6, 531-537.
Kolisoja, P. and Mkel, E. (2001). Instrumentation and mechanical modeling of a full-scale

railway embankment, Proceedings of the 15th International Conference on Soil Mechanics
and Geotechnical Engineering, Istanbul, August 2001. Vol. 3: 2111-2114.
Kolisoja, P., Jrvenp, I., Mkel, E., and Levomki, M. (2000) Instrumentation and model-
ling of track structure, 250 kN and 300 kN axle loads, Finnish Rail Administration, Publica-
tion A 10/2000, 99 pp (in Finnish).
Kopf, F. and Adam, D. (2005). Dynamic effects due to moving loads on tracks for high-speed
railways and on tracks for metro lines. Proceedings of the 16th International Conference
on Soil Mechanics and Geotechnical Engineering, Osaka, September, 2005. Vol. 3: 1735
1740.
Li, D. and Davis, D. (2005). Transition of Railroad Bridge Approaches. Journal of Geotechni-
cal and Geoenvironmental Engineering, 131, no. 11, 1392-1398.
Lundqvist, A. and Dahlberg, T. (2004). Dynamic train/track interaction including model for
track settlement evolvement. Progress report, March 2, The SUPERTRACK project,
Linkping University, Sweden.
Madshus, C., Lacasse, S., Kaynia, A. and Hrvik, L. (2004). Geodynamic Challenges in High
Speed Railway Projects. Geotechnical Engineering for Transportation Projects, ASCE
GSP 126, pp. 192-215.
Olofsson, S-O. & Hakami, E. (2000). Interaction Spr-Bro-Grund-Jord, Sttningar i vergng

mellan bro och till fartsbank fr lanserade broar. (in Swedish) Itasca Geomekanik, Bor-
lnge.
Methods of Analysis of Damaged
Masonry Arch Bridges
PRIORITY 6
INTEGRATED PROJECT

Author/s: J. Bie, T. Kamiski, P. Rawa, WUT
Revision date:

Dissemination Level
PU Public X
Table of Contents
Summary .................................................................................................................... 5
1 Introduction............................................................................................................. 6
1.1 Subject of the study ....................................................................................... 6
1.2 Practical application of the study.................................................................... 6
1.3 Applied terminology ....................................................................................... 6
1.4 Applied notation ............................................................................................. 8
2 Taxonomy of defects in masonry arch bridges ....................................................... 9
2.1 General conception........................................................................................ 9
2.2 Hierarchical classification of masonry bridge defects..................................... 9
2.3 Defects and structure condition.................................................................... 11
3 Methods applied in analysis of damaged masonry arch bridges .......................... 12
3.1 Classification of methods for analysis of damaged masonry arch bridges... 12
3.2 Analytical methods....................................................................................... 14
3.2.1 Introduction ....................................................................................... 14
3.2.2 The maximum stress analyses.......................................................... 14
3.2.3 The thrust line analyses .................................................................... 18
3.2.4 The mechanism methods.................................................................. 22
3.3 Computer-based applications ...................................................................... 26
3.3.1 Introduction ....................................................................................... 26
3.3.2 Archie-M ........................................................................................... 26
3.3.3 RING................................................................................................. 28
3.3.4 FEM systems .................................................................................... 30
3.3.5 DEM systems.................................................................................... 42
3.3.6 DDA systems .................................................................................... 43
3.3.7 Explicit formula analysis.................................................................... 45
3.4 Semi-empirical methods .............................................................................. 45
3.4.1 Introduction ....................................................................................... 45
3.4.2 The MEXE method............................................................................ 46
3.5 Comparison of the analysis methods ........................................................... 48
4 Conception of advanced analysis of damaged masonry arch bridges by means of
FEM ..................................................................................................................... 50
4.1 General idea ................................................................................................ 50
4.2 Selection of defects types ........................................................................... 50
4.3 Parameters of defects description............................................................... 50
4.3.1 General assumptions ........................................................................ 50

4.3.2 Location () ...................................................................................... 51
4.3.3 Extent () .......................................................................................... 51
4.3.4 Intensity () ........................................................................................ 51
4.4 FE model of masonry bridge structure ......................................................... 51
4.4.1 Basic assumption.............................................................................. 51
4.4.2 Arch barrel ........................................................................................ 52
4.4.3 Backfill .............................................................................................. 54
4.4.4 Loads ................................................................................................ 55
4.5 Procedure of the load carrying capacity assessment................................... 55
4.6 Comparison with other methods .................................................................. 56
4.7 Modelling of selected types of defects ......................................................... 57
4.7.1 Deterioration ..................................................................................... 57
4.7.2 Discontinuity...................................................................................... 61
4.7.3 Loss of material................................................................................. 64
5 Final remarks........................................................................................................ 68
5.1 Conclusions ................................................................................................. 68
5.2 Plans for further works ................................................................................. 68
References ............................................................................................................... 69
Summary
The present study brings up an issue of defects in masonry arch bridges and possi-
bilities of their analysis by means of deterministic analytical methods. Description of
the defects is based on the taxonomy developed within WP3 of this project. Pre-
sented methods used in analysis of masonry arch bridges with defects are classified
in three groups: analytical methods, computer-based applications and semi-empirical
methods. The analytical methods are based on: the maximum stresses, the thrust
line or the mechanism method. For each method detailed description from the point
of view of the defects modelling possibilities is provided. Ways of representation for
various defect types in all the methods is explained and their usefulness is evaluated.
Precise analysis of selected defects influence on the ultimate load for various bridge
geometries is presented with application of the Finite Element Method. Detailed de-
scription of the proposed 2D model of the bridge is included in the report. Methods of
modelling for selected defects in FEM is explained. Some examples of the analyses
and results of the defects influence on the ultimate load are presented in several
diagrams.
General conclusions on application of all presented methods of the defects model-
ling are given. The influence of the main defect types on masonry bridge structures
analysed by means of FEM is commented.
1 Introduction
1 Formal information
This technical report is prepared on the basis of Contract No. TIP3-CT-2003-001653
between the European Community represented by the Commission of the European
Communities and the Skanska Teknik AB contractor acting as Coordinator of the
Consortium.
Presented report is a contribution of Wrocaw University of Technology to the Guide-
line SB-LRA (2007).
1.1 Subject of the study

The present study brings up an issue of defects in masonry arch bridges and possi-
bilities of their analysis by means of deterministic analytical methods. Only defects of
the arch barrel are being taken into account. A matter of concern of the study are
analyses giving in a result the load carrying capacity of damaged structures. Each of
the presented methods is briefly described in a practical way explaining its general
idea and capabilities for introduction of defects into a given model. The defects are
defined on the basis of classification proposed in SB-CAI (2007) including seven
main types. A special attention is paid to an advanced modelling of defects in FEM.
1.2 Practical application of the study

The study will help engineers and bridge inspectors to choose appropriate method for
analysis of damaged bridge structure and to introduce a defect into the chosen
model. The crucial features of the considered methods are explained and its possi-
bilities and limitations for selected cases are discussed. Some suggestions for best
solution choice are given taking into account desirable simplicity from one side and
required complexity from the other side. Detailed explanation for modelling of defects
in FEM is provided what enables independent analysis of damaged bridge structures
to not much advanced engineers.
1.3 Applied terminology

Analytical method a method based on analysis of a mathematical model created
on theoretical bases providing information on the structural behaviour.
Arch barrel a main structural element of a masonry arch bridge comprising each
span, having a curved shape, supporting all loads by compression; made of
bricks or stones and mortar.
Backfill material placed between spandrel walls above extrados supporting the
running surface and transmitting all loads to the arch barrel; made of soil, sand or
gravel.
Bridge condition general measure presenting bridge technical condition and
bridge usability.
Bridge technical condition measure of a difference between current and de-
signed values of bridge technical parameters, e.g. geometry, material characteris-
tics, etc..
Bridge usability measure of a difference between current and designed values of

bridge service parameters, e.g. load capacity, clearance, maximum speed, etc..
Crown a middle part of an arch barrel.
Defect each effect diminishing a bridge condition.
Empirical method a method based on analysis of full-scale physical models cre-
ated on field test results providing information on the structural behaviour.
Extrados the outer (top) surface of an arch barrel.
Field test physical loading and measurement of a real full-scale structure in natu-
ral location and environment.
Intrados the inner (bottom) surface of an arch barrel.
Laboratory test physical loading and measurement of a physical model usually in
reduced scale, in controlled laboratory environment.
Mathematical model a virtual theoretical model of a real structure.
Physical model a real full-scale or reduced-scale structure.
Plastic hinge a mode of a local failure in the arch barrel separating it into two
parts rotating to each other about a hinge arising from cracking and crushing of
material.
Ring a single row of masonry blocks (brick or stones) creating a part or the whole
arch barrel.
Semi-empirical method a method based on analysis of both mathematical and
physical models created on theoretical bases and laboratory or field test results
providing information on the structural behaviour.
Spandrel wall a wall built on the edge of the arch barrel closing and keeping a
backfill.
Springing a part of an arch barrel in the place of the connection with a support.
Structural behaviour information on structure response to given loads describing
distribution of internal forces, deformation, load carrying capacity, a failure mode,
etc.
Theoretical bases a set of theoretical assumptions for structural analysis includ-
ing features of model, predicted behaviour and limit criteria for a failure of a struc-
ture.
Thrust line a set of points along a structural element indicating position of the re-
sultant of inertial forces in every cross-sections.
Virtual structure a model of a structure representing a real one with strictly de-
termined rules, on the series of assumptions using theoretical bases.
1.4 Applied notation
A an area of the arch barrel cross-section,

B a width of the arch barrel,
d a height of the arch barrel cross-section,
dm mortar loss in Archie-M,
e eccentricity of the force resultant in the arch barrel cross-section,
E a modulus of elasticity of the masonry material,
Ej modulus of elasticity of the joint,
Eu modulus of elasticity of the stone/brick unit,
fc compressive strength of the masonry material,
ft tensile strength of the masonry material,
Gf dissipated fracture energy,
h a height of the soil over the arch barrel crown,
Iy a moment of inertia of the arch barrel cross-section,
k a ratio between the maximum strain and the critical strain of the masonry mate-
rial,
L arch barrel span
M bending moment in the cross-section of the arch barrel,
N axial force in the cross-section of the arch barrel,
n number of rings of the arch barrel,
N1 number of segments in the longitudinal direction of the arch barrel,
N2 number of segments in the transverse direction of the arch barrel,
P live load in a form of concentrated force acting on a structure,
t height of the yield block in the cross-section of the arch barrel,
T shear force in the cross-section of the arch barrel,
tj thickness of the mortar joint,
tu thickness of stone/brick unit,
a density of the masonry material,
a ductility of the masonry material,
a coefficient of friction between the arch barrel and soil,
j Poissons ratio of the mortar joint,
u Poissons ratio of the stone/brick unit,
defect impact factor,
2 Taxonomy of defects in masonry arch bridges

2.1 General conception
Proposed classification of defects is based on the one given in the SB report SB-CAI
(2007). In the proposed strategy of defect classification a multi-level hierarchical or-
der of defects is considered (Figure 2.1).
Figure 2.1 Conception of hierarchical classification of defects according to SB-CAI

(2007)
The classification of defects is based on description of their effects i.e. their influence
on the measurable (physical or chemical) features of structural elements, which can
be identified during visual inspections as well as by means of advanced testing
methods.
2.2 Hierarchical classification of masonry bridge defects

Considered classification of defects of masonry bridge structures is consistent with
the strategy presented in chapter 2.1. The defects are divided into 7 main types pre-
sented in Figure 2.2 which have further subtypes given in detail in SB-CAI (2007).
MASONRY STRUCTURE ELEMENT
CONTAMINATION
DEFORMATION
DETERIORATION
DISCONTINUITY
DISPLACEMENT
LOSS OF MATERIAL
Figure 2.2 Main types of masonry bridge defects according to SB-CAI (2007)
Definitions of all defect types and their examples observed in stone and brick bridges
are given in Table 2.1.
Table 2.1: Classification of main types of defects observed on masonry arch bridges
Type of Examples
defect Stone bridge Brick bridge
CONTAMINATION
Appearance of any type
of a dirtiness
or a plant vegetation
DEFORMATION
Geometry changes incompatible
with the project, with changes
of mutual distances
of structure points
DETERIORATION
Degradation of physical
and chemical structural features
DISCONTINUITY
Break in a structure
material continuity
DISPLACEMENT
Displacements of a structure or
its part incompatible with project
but without changes of mutual
distances of structure points
LOSS OF MATERIAL
Decrease of structure
material amount
2.3 Defects and structure condition

Each type of defects has an influence on the technical condition and/or on the load
carrying capacity of the structure. These relations are given in Table 2.2. It is visible
that the technical condition is dependent on every type of a defect but the load carry-
ing is sensitive only to some of them.
Table 2.2 Influence of defect on technical condition and load carrying capacity
Defect Technical condition Load carrying capacity

Contamination
Deformation
Deterioration
Discontinuity
Displacement
Loss of material
In this study only the types of defects influencing the load carrying capacity are being
considered.
3 Methods applied in analysis of damaged masonry arch

bridges
3.1 Classification of methods for analysis of damaged masonry
arch bridges
Structural analysis is a general tem describing the structure behaviour under a given
load condition. It gives sometimes distribution of internal forces, predicts failure mode
or indicates location of plastic hinges, but in some cases it provides only an assess-
ment of the load carrying capacity. The letter type of an output is accessible in most
of analysis approaches and the proposed classification is based on the load carrying
capacity assessment.
The main division of methods for defining the load carrying capacity of a bridge lays
between analysing a mathematical and a physical models of a structure (Figure 3.1).
It gives respectively theoretical and experimental methods. In this study only the
theoretical methods are going to be presented.
M ETHODS OF LOAD CARRYING CAPACITY ASSESSM ENT
M ATHEM ATICAL M ODELS PHYSICAL M ODELS
THEORETICAL LABORATORY FIELD

BASES TESTS TESTS
ANALYTICAL M ETHODS SEM I-EM PIRICAL M ETHODS EM PIRICAL M ETHODS
Figure 3.1 Methods of load carrying capacity assessment

Proposed classification distinguishes methods of analysis in respect of the general
approach. According to this the approaches can be divided into 3 groups: analytical,
semi-empirical and empirical methods. The first group is based on analysis of a
mathematical model created on theoretical bases. Semi-empirical methods are
based on both mathematical and physical model being created on theoretical bases
and laboratory or field test results. Empirical methods are based only on a physical
model and field test results.
There are many various types of analytical methods for analysis of masonry bridge
structures. Selected references (Gilbert and Melbourne 1994, Heyman 1982, Van der
Pluijm 1997, Livesley 1978, Orduna and Lourenco 2005a & 2005b) are presented in
the last chapter of the report. At the first level of the proposed classification a kind of
algorithm criteria to be met by the structure are taken into account. At the lower level
the possible models of geometry and material are considered. To be precise the ana-
lytical methods are related with structural analysis of a bare arch barrel and presence
of the backfill can be taken into account in various ways not strictly determined by a
given method.
The analytical methods can be classified as follows:
The maximum stress analyses:

Elastic arch rib,
Inelastic arch rib,
Elastic spatial frame,
The thrust line analyses:
Line of thrust,
Zone of thrust,
Funicular network,
The mechanism methods:
Rigid blocks,
Rigid-plastic blocks,
Volumetric blocks.
The analytical methods have usually numerical representations in the form of com-
puter-based applications giving mostly discrete solutions. These applications present
a comprehensive approach to analysis of the whole masonry bridge structure giving
a possibility to consider also an interaction of a backfill or even spandrel walls and
abutments.
The most common computer-based applications are:
Archie-M,
RING,
FEM systems,
DEM systems,
DDA systems,
Explicit formula analysis.
These computerized approaches use different analytical methods presented in Table
3.1. Some of them are strictly related to one specific analytical method and other can
be used to various types of theoretical analyses.
Table 3.1 Analytical methods used in computer-based applications
Analytical methods
Computer-based
applications Maximum stress The thrust line The mechanism
analysis analysis method
Archie-M
RING
FEM systems
DEM systems
DDA systems
Explicit formulae
Another approach to analysis of the load carrying capacity of masonry bridges using
theoretical bases comprise semi-empirical methods. However these methods are
also based on experimental techniques like laboratory tests or field tests. They usu-
ally have a form of simple formulae.
An example of a semi-empirical method is:
The MEXE method.
3.2 Analytical methods

3.2.1 Introduction
Available analytical methods applied in analysis of masonry arch barrel are subjected
to one of three general theoretical bases defining features of a model, expected be-
haviour and limit criteria for a failure of a structure. Two of them: maximum stress
analysis and the thrust line analysis are static approaches while the mechanism
method comprise a kinematic approach.
3.2.2 The maximum stress analyses
3.2.2.1 General remarks

This approach is based on determination of the maximum stresses in elements of the
masonry structure and comparison of them with the material strength. It applies con-
tinuum mechanics equations for static calculation of the internal forces. There are
several models used in such approach but all of them consist of one-dimensional bar
element. Differences in methods algorithms applied in this method depend on mate-
rial model.
These methods relate mainly to single-span bridge structures or to multi-span struc-
tures treating each span individually. Besides in its basic form it is only analysing a
bare arch neglecting influence of the backfill. A possible way of considering soil influ-
ence is introducing additional springs modelling lateral earth pressure proposed e.g.
in Martn-Caro et al. (2002).
3.2.2.2 Applied models

Geometry models:
1. Arch rib
Main features:
static scheme of the arch barrel: fixed-end, 2-pin or 3-pin arch rib,
represents a unit width of the arch barrel,
required geometric characteristics: A, Iy, d, B,
2. Spatial frame
Main features:
static scheme of the arch barrels: group of fixed-end arch ribs jointed by
transverse beam elements,
each of the arch ribs represents the related width of the arch barrel,
required geometric characteristics: Ai, Iyi, di, Bi,
Figure 3.2 Geometry models in maximum stress analyses: 3-pin (a, b), 2-pin (c)
and fixed-end arch rib (d) or spatial frame (e)
Material models:
1. Elastic
Main features:
constitutive model: linear elastic,
homogenous,
required material properties: E, fc, ft, ,
2. Inelastic
Main features:
constitutive model: linear elastic-perfectly brittle or linear elastic-perfectly
plastic, no tensile resistant (NTR),
homogenous,
required material properties: E, fc, ,
3.2.2.3 Elastic arch rib or spatial frame

This is a simple approach to analysis of masonry arch spans treating them as an
arch ribs made of an elastic material. These assumptions make the solution proce-
dure a typical problem of a static calculation. Such a method gives reasonable results
in cases when the stresses provoked by loads are relatively low and strongly lower
than the material strength. It can be appropriate e.g. in dynamic analysis.
The selection of a method for solving a given problem depends on static determina-
tion of considered model. If it is statically determinate (Figure 2.2(a)) it can be solved
formulating equilibrium equations for horizontal and vertical forces and bending mo-
ments. If the structure model is statically indeterminate (Figure 2.2(c), (d)) it is solved
generally employing integration of equilibrium equations, force method or displace-
ment method.
In case of 2-pin arch scheme (Figure 2.2(b)) the problem can be reduced to determi-
nation of the horizontal force component at springings from eq. 1:
M P (s )y (s )
EI (s )
ds
H= eq. 1
y (s )
2
EI (s )
ds
As the calculation result the internal forces from applied loads and hence stresses
are obtained. Then the stresses are compared with the remaining material resistance
and in this way the load carrying capacity can be estimated.
An idea of considering plastic hinge formation (especially in fixed-end arch scheme)
is introduction of pins in places where highest stresses are detected.
3.2.2.4 Inelastic arch rib

This method accommodates possibility for taking into account non-linear behaviour of
masonry i.e. no resistance in tension and elastic-perfectly brittle or elastic-perfectly
plastic behaviour in compression. The concept proposed e.g. by Brencich and De
Francesco (2004a) is based on static calculation presented above but additionally
involves incremental procedure for loading and responding modification of the static
scheme of the model during the analysis.
The whole structure is divided into segments and calculations are carried out in the
nodes arising from the division. At every step of the solution procedure an increment
of the load is applied and elastic analysis is performed giving as a result internal
forces in the nodes. Then the height of the effective thickness xi of a section is calcu-
lated from average values of internal forces for both the section nodes. The effective
thickness is established according to the assumed approach i.e. neglecting the ten-
sile zone and/or the crushed area. In the next step the geometry of the arch is up-
dated: the section heights are equal to values of xi and their axes are offset suitably
(Figure 3.3). The connections between ends of adjacent segments are infinitely rigid.
Figure 3.3 Scheme of geometry in elastic-plastic analysis

The possible outcome of such an approach is an approximate presentation of internal
forces and effective thickness of the arch barrel from any load (Figure 3.4). The limit
state is also possible to obtain with probable presentation of plastic hinges and the
load carrying capacity value.
Figure 3.4 Example of distribution of the effective thickness of the arch barrel in the
inelastic method (Brencich et al. 2001)
3.2.2.5 Defect modelling

In respect of significant simplification imposed by the assumed model only simple
manners of modelling defects can be taken into account. Any modifications of both
geometric and material properties can be applied only to the whole cross-section of a
model.
Defects possible to consider are presented in Table 3.2.
Although it is visible in Table 3.2 that most of defect types can be taken into account,
one has to remember that the results received by means of this approach are very
limited. Therefore even exact modelling of defects in this method not always provides
satisfying outcome. However quite easily accessible results giving general informa-
tion on the internal force distribution are in cases of support displacements. Due to
one-dimensional models any modifications of material properties relates to the whole
cross-section of elements.
Table 3.2 Defect modelling in maximum stress analyses

Defect Representation Comment
Deterioration Modification of material properties Applicable only to the whole cross-

(strength (E, fc, ft) section
reduction) Introduction of pins Applicable to fixed-end arches
Discontinuity
(longitudinal Modification of geometric characteristics Applicable only to the whole cross-
crack and (A, Iy, B) section
fracture)
Loss of material Modification of geometric characteristics Applicable only to the whole cross-
(brick/stone) (A, Iy, d) section
Deformation
Modification of static scheme geometry
(deflection)
Displacement
Applicable only to statically indeter-
(translation and Enforced movements of supports
minate structures
rotation)
3.2.3 The thrust line analyses

In this general method location of the thrust line is being analysed. The line of thrust
is a term describing for the set of points along the arch indicating position of the re-
sultant of inertial forces in a given cross-sections. Criterion for the load carrying ca-
pacity of the arch is defined here by limits for the thrust line location. There are few
variants of the method which differs from one another by the size of the limits. Also
some differences between analysis procedures are related to applied material model.
Calculations of the thrust line location can be performed using the equilibrium equa-
tions or by solving a linear programming problem.

Geometry models:
1. Arch rib
Main features:
static scheme of the arch barrel: fixed-end arch rib,
2. Spatial network
Main features:
static scheme of the arch barrels: fixed-end spatial network consisted of bar
elements with geometry adapted to applied load scheme (Figure 3.5),
required geometric characteristics: di,
Figure 3.5 The force network for an arch barrel for the self-weight and a point load
(O'Dwyer 1999)
Material models:
1. Elastic
Main features:
constitutive model: linear elastic, infinite compression strength,
homogenous,
required material properties: E, ,
2. Elastic, NTR
Main features:
constitutive model: linear elastic, no tensile resistant (NTR), infinite
compression strength,
homogenous,
required material properties: E, ,
1. Elastic-plastic
Main features:
constitutive model: non-linear elastic-plastic, no tensile resistant (NTR),
homogenous,
required material properties: E, fc, ,
3.2.3.3 Line of thrust

The analysis controls location and slope of the thrust line within the arch barrel. The
analysed parameter describing location of the line of thrust can be an eccentricity of
the force resultant e being a function of normal force N and bending moment M act-
ing in a considered cross-section. The slope of the thrust line is controlled by means
of a relation between normal force N and shear force T. Depending on the assumed
theory and material model there are various limits for the line of thrust position. Struc-
ture meeting these conditions for given loads is assumed to be safe.
The earliest version of the method is called the middle third rule. It imposes a limita-
tion for the thrust line to lie within a section core which for rectangular cross-section
is equal one-third of the total depth which is given by eq. 2. This criterion has been
arrived at on the basis of elastic theory assuming elastic material.
M d
e= eq. 2
N 6
Such a rigorous limit makes the method extremely conservative and at the same time
is very difficult to be met. It can be satisfied only in cases when the structure has
been properly designed and the dead loads considerably dominates over live loads.
The above-mentioned problem caused that the method was relaxed to the middle
half rule what increased the size of allowed core for the force resultant position to the
half of the cross-section (eq. 3).
M d
e= eq. 3
N 4
The last variant of the method was proposed by Heyman (1982). He concluded that
even if the line of thrust runs outside the allowed core in one section it does not
threaten the safety of the whole structure. A structure will collapse only if the line of
thrust reaches the edges of the arch barrel at least in four sections that would convert
it into mechanism. According to this idea the allowed space for the thrust line is the
whole section of the arch barrel what can be written as:
M d
e= eq. 4
N 2
In the last approach an important assumption is infinite material compression

strength which enables the line of thrust to lie just at the edge of a cross-section. This
presumption is rather not realistic but in most masonry bridge structures mean
stresses are relatively low and the real solution is very near to the one obtained in
this way.
All the variants for the thrust line location can be summarized by employing so called
geometric factor of safety (Heyman 1982). This factor is defined as the ratio be-
tween the actual arch barrel thickness and the minimal thickness of similar arch ac-
commodating the line of thrust from a given set of loads. According to this definition a
structure with a load giving the thrust line satisfying the limit from eq. 2 has a geomet-
ric factor of safety equal to 3.
The slope of the thrust line should meet a condition dependent on a coefficient of
friction between voussoirs of the arch (eq. 5) ensuring that a slip of blocks will not
occur:
N
eq. 5
T
3.2.3.4 Zone of thrust

The zone of thrust method, presented first time by Harvey (2001), generally is very
similar to the line of thrust method but additionally takes into account the fact that the
material compression strength is definitely limited. This assumption slightly modifies
condition of the allowed space for the force resultant in the cross-section of the arch.
It is based on the elastic-plastic material model. The method permits creation of a

rectangular yield block around the point of the force resultant position (Figure 3.6).
According to the method assumptions the force resultant lies in the middle of that
yield area. The height of the yield block t is equal to the minimal one providing trans-
mitting the normal force N at the given material strength fc and an arch barrel width B
what can be written as:
N
t= eq. 6
fc B
Hence the resultant force in every section of the arch can not lie nearer to the edge
of the arch then t/2 (Figure 3.6).
force resultant
t
e
yield d
t block
zone of thrust
Figure 3.6 Yield block

Then the condition for the maximum eccentricity of the resultant force is as follows:
M d t
e= eq. 7
N 2
3.2.3.5 Funicular network

It is a spatial application of the thrust line method proposed by O'Dwyer (1999).
Characteristic feature of that approach is a geometry model - a spatial network -
which is being created suitably to the given load scheme according to the expected
load paths. The whole structure is discretized into elements connected in nodes
(Figure 3.5). All the loads, live as well as dead one, are applied to the structure only
in these points in the form of concentrated forces. The limitations for the thrust line
location, which are controlled in the nodes in every part of the network, can be the
same as in the method variants presented above.
The output of the analysis can be both the collapse load for a given pattern of im-
posed loading or the geometric factor of safety for a set of known loads. Solutions for
these two cases can be obtained by solving equilibrium equations for every node of
the model and employing a linear programming algorithm for maximising a live load
factor or geometric factor of safety respectively to the analysis type.

Types of defects possible to introduce into a model are strictly related to the applied
variant of the thrust line method. E.g. in the funicular network analysis neither mate-
rial properties (E, fc) nor most of the geometric characteristics (Iy, A and B) are not
being taken into account but the only parameter to be controlled is the allowed space
for the line of thrust (e).

Generally the thrust line analysis is appropriate for monitoring the influence of defects
on the route of the line of thrust in the arch barrel profile what is clearly visible in a
graphical form. That is why the method is especially effective in determining the
worst locations of defects in the structure. At the same time it indicates the defect
from among other ones which has the biggest influence on diminishing the load car-
rying capacity and should be repaired first.
Table 3.3 Defect modelling in the thrust line analyses
Applicable only to the whole cross-
Modification of material properties section;
Deterioration (E, fc) Applicable only in zone of thrust
(strength method version
reduction)
Modification of constraints for the thrust
line location (e, t)
Discontinuity Modification of geometric characteristics Applicable only to the whole cross-
(longitudinal (A, Iy, B) section
crack and Decrease of the allowed space for the
fracture) line of thrust (e)
Modification of geometric characteristics Applicable only to the whole cross-
Loss of material (A, Iy, d) section
(brick/stone
and joint) Decrease of the allowed space for the
line of thrust (e)
Deformation
(deflection)
Displacement
Applicable only to statically indeter-
minate structures
rotation)
3.2.4 The mechanism methods

Specific approach to analysis of masonry arches presents group of the mechanism
methods. On the contrary to the previously mentioned lower bound analyses these
methods are based on the upper bound approach and hence are used only in a limit
analysis giving a load carrying capacity and a failure mode of a structure.
The algorithm of the methods applies the kinematic approach. It uses an assumption,
which has been confirmed in numerous experimental tests, that a masonry arch be-
comes a mechanism when at least four plastic hinges appear it the arch barrel
(Heyman 1982). However position of the hinges is unknown and hence it has to be
assumed or calculated.
Figure 3.7 Application of the mechanism method for assumed hinge location
A simplified approach uses an assumed mode of a mechanism where position of
hinges is established as the most probable one. For a concentrated load pattern as in
Figure 3.7 one of the hinges should be introduced under the force (in point B). Re-
maining hinges can be located at the springings (A, D) and in the place providing
creation of a geometrically variable scheme (C). Looking for the best mechanism
mode few various hinge locations should be taken until the minimum load is found.
The solution for a given scheme can be carried out by taking bending moments about
hinge points equal to zero or solving equation of virtual work for all parts of the arch
barrel with self-weights Gi and geometrically imposed displacements i (Figure 3.7,
eq. 8).
G11 + G22 + G33 +P = 0 eq. 8

A strict and efficient calculation for unknown scheme is based on application of a lin-
ear programming algorithm with a target function minimising a live load factor. In this
way the most probable mechanism mode is found automatically.
There are few variants of the analysis differing from each other in the matter of ge-
ometry and material model.

Geometry models:
1. Plane arch
Main features:
scheme of the arch barrel: fixed-end plane arch consisted of plane blocks,
required geometric characteristics: d, B, N1, n
2. Spatial vault
Main features:
scheme of the arch barrels: fixed-end spatial vault consisted of volumetric
blocks,
each block represents a part of the arch barrel,
required geometric characteristics: d, N1, N2, n
a) b)
Figure 3.8 Geometry models in the mechanism method analyses: plane arch (a) and
spatial vault (b)
Material models:
1. Rigid
Main features:
constitutive model: infinite compression strength, no tensile resistant (NTR),
homogenous,
required material properties: ,
2. Rigid-plastic
Main features:
constitutive model: rigid-perfectly plastic, no tensile resistant (NTR),
homogenous,
required material properties: fc, ,
3.2.4.3 Rigid blocks

The rigid block analysis presented by Livesley (1978) models an arch barrel as an
assemblage of rigid plane blocks. The division into these parts is regular but not nec-
essarily as dense as the actual number of units of the original arch. As it was
checked experimentally the number of blocks to obtain a sufficiently exact solution is
about 40.
A fundamental assumption of the method is that adjacent blocks which are infinitely
rigid and have infinite strength, at the collapse can move to each other by sliding or
rotating about one of its edge points (Figure 3.10). Movements of all blocks in the
structure are calculated in that way to use minimal energy for the global deformation
i.e. the minimal factor for the given set of live loads.
Figure 3.9 A 4-hinge mechanism

a) b) c)
Figure 3.10 Feasible relative movements of adjacent blocks: rotation about extrados
(a), about intrados (b) or sliding (c)
3.2.4.4 Rigid-plastic blocks

Some important extension of the rigid block analysis has been done by Gilbert 1998.
Here a compression strength of the masonry material is assumed to be finite what
forces displacement of plastic hinges position. It is related with required contact area
between adjacent elements equal to yielding block of the depth t (eq. 6) ensuring un-
constrained transfer of compression force. Now rotations of adjacent blocks take
place about points offset from the block edges at the distance of t/2 (Figure 3.11).
Yield
Plastic block
Hinge
Figure 3.11 Rotation of adjacent blocks in the rigid-plastic blocks analysis

This modification complicates calculation transforming the linear problem to a non-
linear one but applying simple iterative solution strategy it is possible to obtain results
using linear programming problem (Gilbert 1998).
3.2.4.5 Volumetric blocks

In this the most theoretically advanced mechanism method the arch barrel is mod-
elled as a spatial vault consisted of tree-dimensional blocks representing regular unit
parts of the structure. This idea was originally proposed by Livesley (1992) and de-
veloped later by Orduna and Lourenco (2005a, 2005b). The general assumptions for
the solution algorithm are similar to the ones presented for plane blocks analyses
based on virtual works equations. However the three-dimensional problem is much
more complicated and have not been yet effectively computerised. A big advantage
of such approach is prediction of various failure modes more complex than 4-hinge
mechanism like transverse and local failures or skew arch collapse.

The mechanism method as well as the other approaches gives an opportunity to
model some types of defects. However it enables to consider only those defects
which influence a mode of a mechanism which a structure is converted to while it is
collapsing. Modelling of the strength reduction can be done only in the rigid blocks
method by modification of the compressive strength fc. Practically this condition is
incorporated by imposition of the plastic hinge location in a distance equal t/2 from
the arch barrel edge. At the same time according to eq. 6 value of t is also dependent
on the section width B what means that modelling of parallel cracks and fractures or
losses by reduction of B is realized through modification of t.
Table 3.4 Defect modelling in the mechanism methods
Applicable only to the whole cross-
Modification of material properties section;
Deterioration (fc) Not applicable in rigid blocks
(strength method version
reduction)
Modification of a plastic hinge location Not applicable in rigid blocks
from the arch barrel edge (t/2) method version
Discontinuity Modification of the arch barrel width Applicable only to the whole cross-
(longitudinal (B) section
crack and Modification of a plastic hinge location
fracture) from the arch barrel edge (t/2)
Discontinuity Modification of number of rings Simulates separation of rings;
(delamination) (n) Applicable to the whole arch barrel
Reduction of the arch barrel thickness Applicable only to the whole cross-
Loss of material (d) section;
(brick/stone) Modification of a plastic hinge location
from the arch barrel edge (t/2)
Deformation Based on modification of block ge-
(deflection) ometry without creation of hinges
Displacement
rotation)
3.3 Computer-based applications

3.3.1 Introduction
These methods comprise computer-based representations and solutions of theoreti-
cal approaches to analysis of the masonry arch bridges presented in the previous
chapter. They are strictly specialized ready-to-use computer programs being only
applied to masonry arches as well as the general purpose systems like FEM adapted
to this specific problem. Depending on the used method it is possible to obtain vari-
ous output including the load carrying capacity.
3.3.2 Archie-M

An example of a computer software applied to analysis of masonry arch bridges is
Archie-M developed by Harvey 2001. A theoretical basis of the program is the thrust
line analysis and to be precise it is the zone of thrust approach employing a finite
masonry strength. This application ca be used for modelling multi-span arch bridges
together with supports and backfill.
Calculation are carried out on a static scheme of a 3-pin arch where locations of the
pins is chosen as the most likely one for a given load pattern. In respect of that the
program presents a response of the structure to any value of loads in the form of the
most probable zone of thrust location in the profile of the arch (Figure 3.12). It also
graphically indicates location of the hinges. The condition for the structure to be in
equilibrium is that the zone of thrust exists in the interior of the cross sections along
the whole arch. The limit state giving the load carrying capacity can be obtained in-
creasing the live load value up to the moment when the zone of thrust begins touch-
ing the edge of the arch profile in the fourth point.
Figure 3.12 Analysis of a masonry bridge by means of Archie-M

The program provides a numerical presentation of results for each arch segment in-
cluding internal forces and the zone of thrust position. A very helpful option is an
automatic location of the worst position for a given loading vehicle.
The backfill in Archie-M is a continuous body providing a soil-structure interaction like
dispersal of load, at rest, passive and active earth pressure.

Taking into account defects is possible in the program in a limited way by introducing
general changes to geometry or material properties of whole arch barrels. No local
defects can be modelled. One special function enables consideration of mortar loss
in the arch barrel intrados but always involves whole span (Figure 3.13).
dm
d
d
Figure 3.13 Modelling mortar loss in Archie-M

Table 3.5 Defect modelling in Archie-M

Deterioration
Modification of material properties The same applied to all arch bar-
(strength
(fc) rels
reduction)
Discontinuity
Modification of the arch barrel width The same applied to the whole
(longitudinal crack (B) arch barrel
and fracture)
Loss of material Modification of the arch barrel height The same applied to the whole
(brick/stone) (d) arch barrel
Loss of material Mortar loss modification The same applied to the whole
(joint) (dm) (Figure 3.13) arch barrel
Deformation Possible by applying True shape

Modification of static scheme geometry option in arch barrel geometry defi-
(deflection) nition
3.3.3 RING

This is a computer software created by Gilbert (2001) especially for analysis of ma-
sonry arch bridges. It is based on the mechanism method using the rigid or alterna-
tively the rigid-plastic blocks approach. The program analyses a 2D model of a bridge
structure built of arch barrels, supports and backfill giving possibility for creation of
multi-span bridges (Figure 3.14). A particular feature of the software is an analysis of
multi-ring arch barrels enabling separation of rings.
Figure 3.14 Analysis of a masonry bridge by means of RING

The calculation algorithm employs an efficient linear programming technique for solu-
tion of virtual works equations. As the result of an analysis is obtained the minimal
multiplier for the live load of a given scheme triggering formation of a mechanism.
What is more the program provides presentation of a failure mode in the form of a
hinge or sliding mechanism (4 hinges or 3 sliding surfaces for single-span and more
complex mechanisms for multi-span structures). Exact locations of hinges or sliding
surfaces of the limit state are indicated and vectors of the mobilised soil pressure are
graphically presented (Figure 3.14).
A proposed backfill influence is taken into account as a stabilising presence of a me-

dium with self-weight exerting both active and passive lateral earth pressure and dis-
persing loads.

RING software is especially useful in analysis of arch barrels with separated rings
(Figure 3.15). However some additional types of defects are possible to apply but
only those ones which have a global range in a given arch (current version 1.5). A
new 2.0 version of the program planned to be finalized in the nearest future will also
provide an opportunity to model local defects to the structure involving only selected
blocks of arch barrels.
Figure 3.15 Modelling separated rings in RING

Table 3.6 Defect modelling in RING
The same applied to the whole struc-
Deterioration ture (ver. 1.5);
Modification of material properties
(strength
(fc, r, t) Individual values given to selected
reduction)
blocks (ver. 2.0)
Discontinuity The same applied to the whole arch
(longitudinal Modification of the arch barrel width barrel (ver. 1.5);
crack and (B) Individual values given to selected
fracture) blocks (ver. 2.0)
Discontinuity Modification of number of rings Simulates separation of rings;
(delamination) (n) (Figure 3.15) Applicable to the whole arch barrel
The same applied to the whole arch
Loss of material Modification of the arch barrel height barrel (ver. 1.5);
(brick/stone) (d) Various values given to individual
blocks (ver. 2.0)
Possible by applying User defined
option in arch barrel geometry defini-
Deformation
Modification of static scheme geometry tion;
(deflection)
Based on modification of block ge-
ometry without creation of hinges
Displacement
(translation and Enforced movements of supports Accessible only in ver. 2.0
rotation)
3.3.4 FEM systems

Finite element method (FEM) is a general purpose technique appropriate for analysis
of any types of structure including masonry arch bridges. The examples of most
popular computer FEM systems used to analysis of masonry structures are: ANSYS,
ABAQUS or DIANA often with self-implemented user codes destined to these spe-
cific application. Depending on the required precision level and accessible work re-
sources the finite element model of a masonry arch can be made of one-, two- or
tree-dimensional elements. More complex modelling approaches are usually indis-
pensable for considering some types of defects but on the other hand sometimes
increasing the dimension of elements does not gives any additional information. That
is why the right selection of the modelling technique is a crucial problem.
Variety of approaches is also related with different material models that can be ap-
plied in the analysis. FEM gives a lot of possibilities in this matter enabling very de-
tailed material behaviour definition in both pre- and postfailure state.

Geometry models:
1. 1D elements - bars
Main features:
static scheme of the arch barrel: fixed-end arch rib,
2. 2D elements - planes
Main features:
static scheme of the arch barrels: fixed-end plane arch,
represents a profile of a unit width of the arch barrel,
required geometric characteristics: Bi, node coordinates,
3. 3D elements - solids
Main features:
static scheme of the arch barrels: fixed-end solid vault,
represents the whole arch barrel,
required geometric characteristics: node coordinates
Material models:
1. Elastic
Main features:
constitutive model: linear elastic,
homogenous,
required material properties: E, fc, ft, ,
2. Elastic, NTR
Main features:
constitutive model: linear elastic, no tensile resistant (NTR), infinite
compression strength,
homogenous,
required material properties: E,
3. Elastic-plastic
Main features:
constitutive model: non-linear elastic-plastic, limited compressive and ten-
sile strength,
inhomogeneous,
required material properties: (), fc, ft, ,
4. Elastic-plastic, NTR
Main features:
constitutive model: linear elastic-perfectly plastic, no tensile resistant (NTR),
finite compression strength,
homogenous,
required material properties: E, fc,
5. Elastic-brittle, NTR
Main features:
constitutive model: linear elastic-perfectly brittle, no tensile resistant (NTR),
finite compression strength,
homogenous,
required material properties: E, fc,
3.3.4.3 1D modelling
There are various 1D modelling techniques dependent on the applied material model
and related analysis algorithm.
The simplest approach is based on elastic material model and comprise the numeri-
cal solution of the elastic maximum stress analysis presented in chapter 3.2.1. How-
ever the most commonly applied model is a fixed-end arch rib. Pinned models which
impose zero bending moments in pins are usually neglected in that kind of analysis.
The arch is modelled as a set of one-dimensional straight beam elements with re-
quired geometric and material properties. In the analysis the internal forces and
stresses are calculated and compared than with the material strength values.
Figure 3.16 Analysis by means of FEM using 1D modelling

A more reliable behaviour of a masonry arch can be obtained employing the elastic-
plastic NTR material incorporating cracking and yielding proposed by Brencich and
De Francesco (2004a, 2004b). This FEM application corresponds to one of the ine-
lastic approaches presented in chapter 3.2.1. The incremental algorithm can be im-
plemented into a F.E. code. It can be applied to multi-span structures modelling piers
as 1D columns divided into segments analogically to arch barrel elements.
The most important material parameter required in 1D analysis is modulus of elastic-
ity E. Assuming that Poissons ratio of both unit and mortar are equal the average
value of the modulus of elasticity for composite masonry material can be derived
from eq. 9:
E j E u (t j + t u )
E= eq. 9
(E j t u + E u t j )
where:
Ej - modulus of elasticity of the joint,
Eu - modulus of elasticity of the stone/brick unit,
tj - thickness of the mortar joint,
tu - thickness of stone/brick unit
The output of the analysis presents updated geometry equivalent to compressed ar-
eas of the structure, deformation, location of plastic hinges and the order of their ap-
pearance (Figure 3.17). The iterative procedure stops when a load increment possi-
ble to apply is smaller then the assumed tolerance what gives the load carrying ca-
pacity value.
1
4
2 6
7 5
3
Figure 3.17 Elastic-plastic analysis using 1D modelling in FEM system
A slightly different approach was presented by Choo et al. (1991) using tapered
beam elements and elastic-brittle material. Here both cracked in tension and crushed
in compression areas of the arch barrel section are removed for calculation in the
next step of an incremental procedure. The axis of an updated arch geometry is dis-
placed to the central line of the current effective arch (Figure 3.18). The effective
thickness is calculated in every node and varies linearly between them on each sec-
tion length.
Figure 3.18 1D modelling geometry in elastic-brittle analysis (Choo et al. 1991)

The presence of backfill in 1D models can be taken into account as a distributed ex-
ternal loading equal to its self-weight and live load dispersed by the soil along the
arch. A stabilising effect of the backfill like passive earth pressure can be modelled
by application of additional vertical and horizontal or radial springs.

Modelling of defects in 1D FEM is agreeable with strategy presented for analytical
method based on the maximum stress analysis.
Table 3.7 Defect modelling in 1D FEM
Deterioration Modification of material properties Applicable only to the whole cross-

(strength (E, fc, ft) section
reduction) Introduction of pins
Discontinuity Modification of geometric characteristics Applicable only to the whole cross-
(longitudinal (A, Iy, B) section
crack and imposed effective thickness of the arch Applicable in inelastic arch rib
fracture) barrel sections method
Modification of geometric characteristics Applicable only to the whole cross-
Loss of material (A, Iy, d) section
(brick/stone) Imposed effective thickness of the arch Applicable in inelastic arch rib
barrel sections method
Deformation
(deflection, slip)
Displacement
Enforced translation or rotation of sup-
(translation and ports
rotation)
Two-dimensional model of a masonry arch bridge is usually created as a vertically
situated plane arch representing a profile of a unit width of the arch barrel. There are
several modelling techniques with respect to the homogeneity of the arch material.
Figure 3.19 shows the applied methods differing from each other in the kind and
boundaries of the employed materials. In the first approach (a) the whole arch barrel
is modelled with a homogenous material what is called a macro-modelling. Another
methods (b) and (c) a simplified micro-modelling include a division into discrete
segments acting between one another through the contact surfaces. It enables de-
tachment of adjacent parts simulating cracking. The next techniques (d)-(f) incorpo-
rate additional joint elements of a different material properties modelling selected
joints of a masonry what can be called a detailed micro-modelling. Depending on
the required level of accuracy and the expected failure mode there are modelled ra-
dial joints only (d), tangential joints only (e) or both of them (f). In the approaches
predicting division into masonry blocks these segments include a piece of masonry
consisted of few units. Such a method is detailed enough to give reasonable results
with regard to a failure mode and the load carrying capacity value in comparison with
the experimental tests. More fine mesh e.g. responding to individual masonry units,
does not bring much benefits but significantly complicates and prolongs the solution.
Figure 3.19 Applied techniques of 2D modelling in FEM

In respect of a relatively much bigger dimension of a masonry arch span in the trans-
verse direction than the arch barrel thickness, two-dimensional numerical models
assume usually plain strain condition.
The behaviour of masonry is dominated by the response of its weakest part which
comprises mortar. It is confirmed by numerous tests where any failures like cracking,
crushing or sliding appear usually in this area. This fact justifies a common simplifica-
tion to model masonry as a plain concrete material. It is applied to the whole struc-
ture for macro-model or to joint elements only in macro-model; the remaining parts -
masonry blocks can behave linear-elastic. This concept implicates isotropy of the
material in the frame of each element.
The general description of a 2D nonlinear constitutive model of a concrete-like ma-
sonry consists of three elements (Figure 3.20):
prefailure behaviour - () relation,
limit conditions (limit domain),

postfailure behaviour - yielding, softening, cracking or crushing.
Due to such a complex material model of at least one of a structure part, the solution
algorithm must be incremental and nonlinear. Depending on the approach the consti-
tutive relations can be more or less complicated. The material behaviour in all the
stages is variously defined for tension and compression what is described by uniaxial
- relations shown in Figure 3.20.
The prefailure behaviour in the simplest approach is linear elastic for both compres-
sion and tension. In more sophisticated model at the higher compressive stress level
it becomes nonlinear (Figure 3.20). For the linear range of the material behaviour the
modulus of elasticity E considering Poissons ratios of unit and mortar can be derived
from eq. 10:
1
E=
u E j j Eu j eq. 10
u + j + 2u j u2
E u E j j (1 u )E j + u (1 j )E u E j E u

2

where:
tj tu
j = , u = ,
t j + tu t j + tu
j - Poissons ratio of the mortar joint,

u - Poissons ratio of the stone/brick unit,
Figure 3.20 Concrete-like material - relationship in pre- and postfailure states

There are various limit domains for concrete-like material (proposed e.g. by Ali and
Page (1988), Loo and Yang (1991), Hibbit et al. (2005)) but all of them have similar
shapes (presented in Figure 3.21) assuming von Mises based envelope in compres-
sion and considerably limited tensile stresses being about 1/10 of the maximum
compressive ones.
Figure 3.21 Limit domains in 2D representation

After exceeding the limit boundary by stresses the material starts to behave in accor-
dance with assumed postfailure conditions. The uniaxial behaviour in compression as
well as in tension can be modelled (Figure 3.22) as perfectly brittle (a), perfectly plas-
tic (b) or assuming strain softening (c) (Loo and Yang 1991).
Figure 3.22 Post failure uniaxial behaviour of concrete like material: perfectly brittle
(a), perfectly plastic (b) or strain softening (c) (Loo and Yang 1991)
The perfectly brittle model is practically inaccessible due to difficulties in solution
convergence reaching the limit envelope. On the other hand the perfectly plastic ap-
proach, the most easily attainable one from the numerical point of view, leads to
overestimation of the actual global limit load for the structure but is sufficient in de-
termining the failure mode. The most truthful imitation of the real masonry material
provides the model employing the strain softening law. It is appropriate especially in
tension but can be also applied in compression. Stresses after reaching the critical
value cr are gradually decreasing down to total loss of the material strength (Figure
3.22(c)). The slope of the descending branch should be taken so to give the maxi-
mum strain about 36 times of the critical value (k = 36). The softening function in
the form of -w relationship where w is a crack opening can be linear or nonlinear
providing a constant area under the -w curve being equal to dissipated fracture en-
ergy Gf (Hillerborg 1976).
Figure 3.23 -w relationship for cracking concrete-like material

A backfill can be modelled by means of additional 2D elements providing transfer of

loads and passive reaction on the arch barrel. The material of the soil is usually
nonlinear defined by Mohr-Coloumb or Drucker-Prager limit criteria but also a crude
approach involving linear elastic material is allowed. In the simplest case the backfill
elements are connected with the arch barrel through common nodes. A better ap-
proach introduces a contact surface between arch barrel extrados and the backfill
what enables relative transverse displacements of these parts which react to each
other by friction. In this approach separate elements which do not have common
nodes are required. Another possibility is to model at the soil-structure boundary nar-
row interface elements with low stiffness and strength providing large deformation in
this area (Ford et al. 2003).
Depending on the applied material approximation 2D modelling enables obtaining
various information on the masonry arch behaviour. Using the homogenous continu-
ous model it is possible to observe a general response of the structure with stresses
concentration, their approximate values, cracks lacation and the load carrying capac-
ity (Figure 3.24). Observing the cracks occurrence it is possible to indicate approxi-
mately the potential plastic hinge location in the zones of the cracking concentration
suggested by circles in Figure 3.24(b).
a)
b)
Figure 3.24 Output of 2D modelling in FEM: stress distribution (a)

and cracks location (b)
Also it is possible to determine an approximate thrust line location by calculating the
components of nodal force in the tangential directions in a given cross-section. The
bending moment which is calculated by multiplying each tangential nodal force by its
lever arm is divided by the resultant tangential force being a sum of the nodal tangen-
tial components what gives a lever arm of the thrust line in the section.
More complex models provide furthermore quite good reflection of deformation and
dependently on the chosen modelling technique cracks location and the failure mode
presentation. An example of such an advanced modelling conception was proposed
by Ford et al. (2003).
Figure 3.25 Example of a detailed micro-model in FEM (Ford at al. 2003)

Here a detailed micro-model is used: the masonry arch is divided into masonry
blocks and joint elements of a different properties (Figure 3.25). The letter elements
have a low tensile strength and after exceeding it they lose their stiffness what is
visible by large deformation. This technique enables precise determination of plastic
hinges in those joint elements where the stresses exceeded the limit values (Figure
3.26).
Figure 3.26 Plastic hinge formation in a detailed micro-model in FEM

(Ford at al. 2003)

Incorporation of defects into 2D FEM model, in spite of the limitation to a plain prob-
lem, provides great possibilities and will be more precisely described in
chapter 4.

Modification of material properties Applicable to any selected part on
Deterioration (E, fc, ft) the cross-section height
(strength
reduction) modification of the strain softening law Applicable only in elastic-plastic ma-
(k ,Gf) terial model (details in chapter 4.)
Discontinuity
Modification of the arch barrel width
(longitudinal crack (details in chapter 4.)
(B)
and fracture)
Discontinuity Modelling separate rings By application of longitudinal con-
(delamination) (n) tact surfaces
Discontinuity
Local limits for transfer of tensile forces
(transverse crack in joint elements
and fracture)
Loss in material
Modification of the arch barrel height
(brick/stone (details in chapter 4.)
(d)
and joint)
Deformation
Modification of static scheme geometry (details in chapter 4.)
(deflection)
Displacement
(translation (details in chapter 4.)
ports
and rotation)
The most advanced but at the same time the most demanding approach to analysis
of masonry arch bridges provides three-dimensional modelling in FEM. The main ad-
vantage of this method is an opportunity to model a whole structure including span-
drel walls and/or abutment wings. With respect to such a detailed computer model
also a scope of accessible results is more extensive and precise. Contrary to the
simpler models it is possible to consider specific structural behaviour of a bridge in-
volving transverse effects and analyse different or more localized failure types. A new
feature of such a technique is the transverse dimension of the model enabling taking
into account both various (not only uniform) configuration of loads and variable re-
sponse of the structure in this transverse direction like variable stress distribution.
Figure 3.27 3D models of masonry arch bridges (Frunzio et al. 2001)

Creation of 3D model requires application of volumetric elements defined geometri-
cally only by the coordinates of their nodes. Due to frequently large size of three-
dimensional problems in cases of transversely symmetric bridges it is possible to
model only half of the structure subjected to symmetrical loads.
Figure 3.28 Model of half of a symmetric masonry bridge structure (Fanning &
Boothby 2001)
In comparison with 2D models here the dominant modelling technique is based on
the homogenized composite material not dividing an arch barrel into masonry blocks.
The material behaviour in pre- and post failure states is described in an analogical
way as in 2D models. An obvious difference concerns definition of a limit domain in a
three-dimensional space. The most simplified idea is given by Rankine type criterion
but more refined and more appropriate for concrete-like materials limit surface is de-
termined by Willam-Warnke criterion (Figure 3.29).
Figure 3.29 Willam-Warnke limit domain

A backfill is modelled by means of 3D elements made of linear elastic or inelastic ma-
terial subjected to Mohr-Coloumb or Drucker-Prager limit criteria. It enables transfer
of loads in longitudinal and transverse directions. The modelling techniques for the
soil-structure interaction can be analogical to 2D models i.e. assuming a fixed con-
nection, predicting interface elements or defining contact surface.
The results provided by application of 3D modelling involve three-dimensional image
of stress distribution in the whole structure or its deformation (Figure 3.30). The main
advantage of that kind of analysis is information on the structural behaviour in the
transverse direction. Therefore also additional modes of failure are possible to predict
like punching-through or longitudinal cracking of the arch barrel or spandrel walls
separation.
Application of 3D models is deemed to be especially valuable and reliable in cases of
skew arches providing specific and more complex failure mode.
a) b)
Figure 3.30 Accessible outcome of 3D analysis in FEM: stress distribution (a), defor-
mation (b) (Frunzio et al. 2001, Fanning & Boothby 2001)

Generally modelling of defects in 3D FEM gives a possibility to model each defect
precisely but usually is very time-consuming. Using of this technique is a matter for
choice between an expenditure of work and precision of potential results which
sometimes can be reached by means of simpler methods.

Modification of material properties Applicable to any selected part of a
Deterioration (E, fc, ft) structure
(strength
reduction) modification of the strain softening law Applicable only in elastic-plastic mate-
(k ,Gf) rial model
Discontinuity
(longitudinal Involves application of longitudinal
Introduction of planes of discontinuity
crack vertical contact surfaces
and fracture)
Discontinuity Modelling separate rings Involves application of longitudinal
(delamination) (n) horizontal contact surfaces
Discontinuity Involves application of transverse

Introduction of planes of discontinuity
(transverse vertical contact surfaces
crack and Local limits for transfer of tensile forces
fracture) in joint elements
Losses
(brick/stone (d)
and joint)
Deformation
(deflection, Modification of static scheme geometry
slip)
Displacement
(translation ports
and rotation)
3.3.5 DEM systems

Another approach to analysis of masonry bridge structures enables the Discrete
Element Method (DEM) (Azevedo and Sincraian 2001). It is especially suitable for
problems in which a significant part of the deformation is accounted for by relative
motion between blocks like masonry at the collapse state. Here the structure is usu-
ally represented by 2D distinct, rigid as well as deformable blocks which can displace
independently from one another and interact at contacts between adjacent blocks
(Figure 3.31).
Figure 3.31 Example of Discrete Element Method application

In the Discrete Element Method the structural analysis is based on explicit algorithms
related to solving the dynamic equations of equilibrium between blocks. An important
fact which has to be taken into account here is that the geometry of the system as
well as the number and type of contacts between the blocks, may change during the
analysis.
An output of such an analysis encompasses presentation of stresses distribution,
progressive failure mode associated with cracks propagation and maximal load pos-
sible to apply to the structure.

Table 3.10 Defect modelling in DEM
(strength
(k ,Gf) terial model
Discontinuity
(longitudinal Modification of the arch barrel width
crack (B)
and fracture)
Discontinuity Modelling separate rings By applying boundaries along the
(delamination) (n) longitudinal joints
Discontinuity
(transverse By applying boundaries along the
Modelling separate masonry blocks
crack radial joints
and fracture)
Loss in material
(brick/stone (d)
and joint)
Deformation
(deflection)
Displacement
(translation ports
and rotation)
3.3.6 DDA systems

Discontinuous Deformation Analysis (DDA) is a method based on a system of dis-
crete, discontinuous 2D blocks which are being analysed considering simultaneously
their deformations and rigid body movements (Thavalingam et al. 2001). The compu-
tational algorithm uses the usual process of minimisation of the potential energy.
Elements of a structure can be modelled by means of deformable blocks of any
shape e.g. quadrilateral elements for the arch barrel and polygonal elements for the
fill (Figure 3.32).
Figure 3.32 Example of Discontinuous Deformation Analysis application

In that kind of analysis it is possible to obtain the distribution of internal forces, poten-
tial collapse mechanism and the maximal value of the live load carried by the struc-
ture. The distribution of stresses within the structure has a resolution of the mesh
density due to the method framework imposing constant stress state over each block.

Table 3.11 Defect modelling in DDA
(strength
(k ,Gf) terial model
Discontinuity
(longitudinal Modification of the arch barrel width
crack and (B)
fracture)
Discontinuity Modelling separate rings By application of boundaries along
(delamination) (n) the longitudinal joints
Discontinuity
(transverse By application of boundaries along
Modelling separate masonry blocks
crack and the radial joints
fracture)
Loss of material
(brick/stone (d)
and joint)
Deformation
(deflection)
Displacement
(translation ports
and rotation)
3.3.7 Explicit formula analysis

There are numerous simplified formulae for calculating a load carrying capacity of
masonry arch bridges using other computer-based applications. One of them was
proposed by Martn-Caro et al. (2004). The results being a basis for formulation of
the explicit formula of the method were obtained by means of a FEM analysis of ma-
sonry arch bridges. Authors of the formula carried out extensive parametric studies of
a few hundred cases of masonry bridge structures with various geometric and mate-
rial properties. They analysed structures separately under concentrated and uni-
formly distributed load patterns and hence they presented different expressions for
the ultimate point load Pult (eq. 11) and uniform load qult.
Pult = (AL2 + BL + C)K1K2 eq. 11
where A, B and C are coefficients depending on arch barrel depth to span ratio, K1 is
a coefficient depending on rise to span ratio and K2 is a coefficient depending on
height of fill at crown. Ultimate uniform load qult is given a series of linear equations
dependent on masonry compression strength and span (Martn-Caro et al. 2004).

In that kind of methods possibilities of taking into account any defects is usually very
limited. In the presented method only the global defects with not specified location
are applicable. According to the proposed approach the geometric parameters influ-
ence only the ultimate point load Pult contrary to the material strength which affects
only uniform load qult.
Table 3.12 Defect modelling in the explicit formula analysis
Deterioration Applicable to the whole arch barrel;

Modification of material properties
(strength Accessible by using appropriate
reduction) (fc)
formula
Loss of material
(brick/stone (d)
and joint)
3.4 Semi-empirical methods

3.4.1 Introduction
Semi-empirical methods are a separate group presenting an easy approach to analy-
sis of masonry arch bridges using theoretical bases and experimental techniques.
The output obtained by means of that kind of analysis is very limited and usually
gives only approximate value of the load carrying capacity of a structure. This value
is calculated from a simple equation formulated on a basis of analytical solutions
and/or experimental results (obtained for a large number of cases with different val-
ues of input parameters) and assembled on some theoretical assumptions.
3.4.2 The MEXE method

The most popular semi-empirical method called MEXE gives as an outcome the
maximum permissible axle and linear loads for a structure (UIC 1994, Department of
Transport 2001). It was worked out on a basis of analytical calculation employing 2-
pin parabolic arch subjected to a concentrated load in the middle of span and some
experimental result from full-scale tests. This combined knowledge of arch behaviour
was consolidated and led to creation of diagrams presenting relationship between
arch barrel thickness at the crown d, inside span L, the depth of the fill above the
crown h and idealized axle load Q. Later this relations were described with a simple
function:
740 (d + h )
2
Q= [kN ] eq. 12
L1,3
This is an idealized load what means it is valid for structures satisfying an assump-
tion that the arch span has a parabolic shape and is in the ideal condition. However
the method provides also the load carrying capacity for differently shaped and dete-
riorated structures. It is being realized by applying few modification factors to the
value Q related with profile shape, material type, joint condition and general condition
of a structure, number of spans and dynamic influences (Department of Transport
2001).
The linear load according to Department of Transport (2001). is calculated from eq.
13:
Q kN
q= eq. 13
1,5 m
Besides there are general geometric limitations for application of MEXE:

an inside span L 20 m,
a rise at the crown rc 1/4 of the inside span L,
a depth of fill 30 cm h 105 cm between the upper surface of the arch crown
and the underside of the sleepers,

The MEXE method enables to take into account many types of defects but not in a
precise way. It gives merely approximate values or ranges for modification factors
which can be taken quite freely. Besides only global and not exactly defined defects
are considered.
Defects possible to consider are presented in Table 3.13 (UIC 1994, Department of
Transport 2001).
Table 3.13 Defect modelling in the MEXE method

Modification of a barrel factor: Masonry of any kind in poor condition (many
Deterioration voussoirs flaking or badly spalling, shearing
Fb = 0,7 etc)
(strength
reduction) Modification of a mortar factor:
Loose or friable mortar
Fmo = 0,9
Modification of a condition factor: longitudinal cracks due to differential settle-
Fcm = 0,40,6 ment in the abutments
Discontinuity
Longitudinal cracking found outside of the
(longitudinal Fcm = 0,95 middle third of the barrel, with lengths up to
crack and one-tenth of the span
fracture)
Longitudinal cracking within the middle third
Fcm = 0,85 of the barrel, with lengths up to one-tenth of
the span
Lateral cracks or permanent deformation of
Discontinuity the arch which may be caused by partial
Fcm = 0,60,8 failure of the arch or movement at the abut-
(transverse
ments
crack and
fracture) Cracks in the joint of the extrados between
Fcm = 0,8 sleeper walls and the extrados of the arch
Discontinuity
Fcm = 0,30,7 Diagonal cracks
(skew)
Reduction of the load carrying ca-
For a 4-ring arch barrel, where N is a number
pacity due to ring separation by ap- of separated rings
Discontinuity
plying coefficients: R4-rings = 1 0,2n
(delamination)
For a 6-ring arch barrel, where N is a number
R6-rings = 1 0,146N of separated rings
Loss of material Reduction of the arch barrel thick-

(brick/stone) ness (d)
Modification of a depth factor: unpointed joints, pointing in poor condition

and joints with up to 12.5mm from the edge
Fd = 0,9 insufficiently filled
Loss of material Joints with from 12.5mm to one tenth of the
Fd = 0,8 thickness of the barrel insufficiently filled
(joint)
2
d d j Joints insufficiently filled for more than one
Fd = tenth the thickness of the barrel, where dj
d depth of missing mortar in joint
Modification of profile factor:

Deformation 0,6 Where rc is a rise at the crown and rq is a rise
rc rq
(deflection) Fp = 2,3 at the quarter points
rc
Modification of profile factor: longitudinal cracks due to differential settle-
Fcm = 0,40,6 ment in the abutments
Displacement
(translation and Lateral cracks or permanent deformation of
rotation) the arch which may be caused by partial
Fcm = 0,60,8 failure of the arch or movement at the abut-
ments
3.5 Comparison of the analysis methods

A very important element of the process of modelling the damaged structures is a
selection of the most efficient method. This selection should consider all required as-
pects for the analysed defect type but at the same time simplify the problem as much
as possible.
Influence of each defect can be modelled by many methods but usually some of
them are better and especially appropriate for a given type of a defect. Efficiency of a
method depends on possibility of reliable defect representation as well as useful re-
sults of analysis. Besides its output has to provide a noticeable difference between
damaged and undamaged structure and be sensitive to variations of defect parame-
ters.
Another aspect which would be important from a maintenance point of view is infor-
mation on potential ways and effectiveness of a repair of a structure with a given de-
fect. Peculiar application to designing of strengthening would have a method indicat-
ing the most proper location for reinforcing elements and showing their potential in-
fluence.
Some suggestions on selection of an appropriate approach for modelling various de-
fect types influencing the load carrying capacity are given in Table 3.14.
Table 3.14: Application of various approaches to analysis of masonry arch bridges with defects
Defect
Deterioration Discontinuity Loss of material Deformation Displacement
Approach longitudinal transverse
strength translation and
crack and crack and delamination b rick/stone joint deflection
reduction rotation
fracture fracture
Sustainable Bridges
Theoretical bases
Maximum stress analyses
Elastic arch rib z z
Inelastic arch rib z z z
Elastic spatial frame z z z
The thrust line analyses
SB-4.7.3
Line of thrust z z z
Zone of thrust z z z
Funicular network z z z
The mechanism methods
Rigid blocks z z z
Rigid-plastic blocks z z z
Volumetric blocks z z z z
Computer-based applications
Archie-M z
RING
ver. 1.5 z
ver. 2.0 z z z z
FEM
1D z z
2D z z z z z z z z
3D z z z z z z z
2007-11-30
DEM z z z
DDA z z z
Explicit formulae
Semi-empirical method
MEXE
49 (71)
z - application suggested, } - applicable but not suggested, - not applicable

4 Conception of advanced analysis of damaged masonry

arch bridges by means of FEM
4.1 General idea
The proposed idea of modelling and analysis of damaged masonry arch bridges in-
volves application of FEM. This approach is based on evaluation of defects impact
on the load carrying capacity of structures in a strictly determined procedure pre-
sented below. Only defects influencing the load carrying capacity are analysed and
for each of them individual defect description is presented. The model destined for
incorporation of defects and applied in analyses is the same for all defect types.
Techniques of defect modelling are explained in detail for each of the defect types.
4.2 Selection of defects types

For advanced analysis in FEM the main defect types influencing the load carrying
capacity (given in chapter 2.3) are selected, which are defined as follows:
deterioration represented by reduction of material strength and modulus
of elasticity,
discontinuity longitudinal fracture (parallel to span free edges) with sepa-
ration of material strip,
loss of material loss of blocks and joints material at the intrados of the
arch barrel,
displacement translation of the arch barrel springing and boundaries of
the adjacent area of the soil modelling dislocation of the abutment.
4.3 Parameters of defects description

4.3.1 General assumptions
Description of each defect apart from indication of its type should also contain a nu-
merical dimensioning. Taking into account that the selected geometry model is two-
dimensional the description should geometrically define the defect severity and local-
isation on the arch barrel profile. It can be realized by means of a few parameters
which are proposed to be:
1. Location () - identification of defect appearance position on the arch barrel
length,
2. Extent () - measure of defect size on the arch barrel length,
3. Intensity () - measure of defect size in the arch barrel cross-section,
Besides some of the types of defects require additional parameters which are defined
further in description of each defect modelling procedure.
This approach is useful for parametric studies involving many cases of defects with
variable parameters as well as for analysis of individual structure with specific de-
fects.
4.3.2 Location ()
Description applied for deterioration, discontinuity and losses:
- value of the ratio between the longitudinal coordinate indicating the central
point of a defect xc to the whole arch barrel length l (Figure 4.1), (eq. 14):
x cd ,c,l
d,c,l = eq. 14
l
4.3.3 Extent ()
- value of the ratio between a defect length l to the whole arch barrel length l
(Figure 4.1), given on the percentage basis (eq. 15):
l d ,c,l
d,c,l = 100% eq. 15
l
4.3.4 Intensity ()
- value of the ratio between a damaged height in the cross-section d to the
designed arch barrel thickness d (Figure 4.1), given on the percentage basis
(eq. 16):
d d , c, l
d,c,l = 100% eq. 16
d
Figure 4.1 Representation of the geometric defect parameters.
4.4 FE model of masonry bridge structure

4.4.1 Basic assumption
Model proposed for analysis of the damaged structures is two-dimensional and
represents
1 meter wide strip of a masonry arch bridge. It includes the arch barrel and the soil
over it called backfill (Figure 4.2). This idea comprises a typical approach to analysis
of masonry bridges (Ali and Page 1988, Loo and Yang 1991, Choo et al. 1991b, Ford
et al. 2003). In the study numerical examples of defects incorporation are presented
on the most common constructions among masonry arch bridges: single-span, rec-
tangular in plane structure with single-ring and segmental arch barrel. However the
same techniques can be applied to any structural types of masonry arch bridges
modelled in the proposed way.
Figure 4.2 FEM model of a masonry arch bridge structure.

A numerical model of a structure consists of a set of parameters comprising geomet-
ric characteristics and material properties (Figure 4.3). In the numerical examples,
presented for each damaged type, typical for railway bridge structures values of the
parameters are considered.
Applied two-dimensional model of structure assumes the plane strain conditions what
is justified by the fact that width of real structures is usually much higher then 1 me-
ter.
Figure 4.3 Geometric and material properties of masonry bridge model.
4.4.2 Arch barrel

The arch barrel which is responsible for bearing all loads comprises the most impor-
tant element of a structure. Therefore a proper modelling of this part of a structure
has an essential relevance to obtaining reliable results. Model considered in numeri-
cal examples represents segmental single-ring arch barrel with rise-to-span ratios
r/L0 equal to 1/4 and 1/2 and span lengths L0 from 5 m up to 15 m. The arch barrel
thicknesses d is taken as a function of span: d = 0.2 L0 .
Discrete model of masonry arch barrel was created by means of a semi micro-
modelling technique (according to chapter 3.3.4.5). It means that the whole profile of
the arch barrel is divided into segments called masonry blocks separated by joint
elements of different properties (Figure 4.4). This idea is similar to approach pre-
sented by Ford et al. (2003) and follows a philosophy of rigid block analysis devel-
oped by Gilbert and Melbourne (1994). Assumed number of segments is equal to 40
what provides good representation of a real structure behaviour. Besides the number
of the blocks has very limited influence on the load carrying capacity what was ana-
lysed by authors. Thickness of joint elements is 1 cm which comprises an average
dimension of real mortar joints. The densities of both masonry blocks b and mortar j
are the same and are equal to 20 kN/m3.
block elements
joint elements
Figure 4.4 FEM model of the arch barrel.
Model of the arch barrel geometry is created of 4-node quadrilateral plane elements
with 2 degrees of freedom in each node. Recommended ratio between element sides
is near to 1, therefore the number of elements on the arch barrel thickness should be
large enough.
Both ends of the arch barrel composed of joint elements are supported by fixed
nodes. It models connection of the arch springings with abutments which are as-
sumed to be rigid.
Considered constitutive models of material are different for masonry blocks and
joints. For simplification of a numerical problem the material of masonry blocks is as-
sumed to be linear-elastic with unconstrained compressive and tensile strengths de-
fined only by modulus of elasticity Eu and Poisons ratio u. Values of these properties
are taken as average values for masonry prism i.e. 10 GPa and 0,16 respectively.
More complex constitutive model of joint elements represents a concrete-like material
behaviour. Thus it is non-linear in compression with compressive strength fc equal 10
MPa and linear in tension with considerably limited tensile strength ft about 100 kPa
(Figure 4.5).
Figure 4.5 Relationship for concrete-like material in the uniaxial behaviour.

Pre-failure linear behaviour of joint material is defined by modulus of elasticity Ej = 1
GPa and Poisons ratio j = 0,16. The post-failure behaviour is determined by Con-
crete Damage Plasticity, Compression Damage and Tension Damage options of
ABAQUS. This model assumes appearance of cracking and crushing of the material
after exceeding tensile or compressive strength respectively. The strain softening in
the post-failure state is defined according to Figure 4.6a (in tension) and Figure 4.6b
(in compression). Such a gradual instead of immediate decrease of stresses is indis-
pensable for a numerical convergence of the solution.
Figure 4.6 Post-failure behaviour of concrete-like material: in tension (a), in compres-

sion (b).
4.4.3 Backfill
Model of the backfill has to be created carefully because it comprises very important
and fully-fledged structural component. It fulfils a dual role in the structural behaviour
of a masonry bridge: provides transfer of loads from the track structure to the arch
barrel and exerts a stabilising influence on deformed arch barrel through a mobilised
passive earth pressure.
Proposed element type for modelling the backfill is a 3-node triangular plane with 2
degrees of freedom in each node.
The applied properties of the soil are typical for most encountered structures. Their
values are as follows: modulus of elasticity Es = 100 MPa, Poisons ratio s = 0,3,
angle of internal friction = 45 and cohesion c = 30 kPa. Limit domain is defined by
Drucker-Prager criterion commonly applied to granular media. The density of the soil
is s = 18 kN/m3.
One of the most important features of the used backfill model is its interaction with
the arch barrel. It is realized by introducing a contact surface between these two
parts of the structure. This interface provides mutual reaction of both contacting ele-
ments, disables their overlapping and enables total separation. Possible relative
transverse displacements of the arch barrel and the backfill is ruled by friction defined
by a value of a coefficient of friction = 0,5.
In numerical examples assumed height of the backfill over the arch barrel crown is
equal to 35 cm what comprises a minimum depth of the railway ballast and is the
most unfavourable case. It is important to include in the model of the backfill some
part of the space of the soil beyond the springings. It gives better representation of
this continuous medium behaviour. Boundary conditions for the backfill at the bottom
edges are fixed supports. Lateral edges of the backfill have only vertical movements
released what simulates continuity of this semi infinite medium beyond these edges.
4.4.4 Loads
Definition and application of loads comprises another crucial issue for analysis of
masonry bridge model. There are generally two different kinds of loads which has to
be distinguished: dead and live loads. Dead load comprise self-weights of each struc-
tural component according to its density. In numerical examples live load is assumed
to be represented by single axle load P of a vehicle standing on the structure. This
axle is acting on the whole width of the model what is obvious in case of two-
dimensional geometry model. In respect of presence of the truck structure the force
is dispersed in longitudinal direction. Therefore the live load is modelled as a uni-
formly distributed load on a length of 50 cm (Figure 4.7).
Another significant issue is the way and order of the loads application. Due to the
model nonlinearity and the loading history dependence all loads must be applied in-
crementally. Therefore in the proposed analysis during the first step the dead loads
are increased up to the final value and only then in the next step the live load starts
to act.
P, u
rigid 50 cm
plate
Figure 4.7 Application of live load to structure.

To monitor the structural behaviour under the live load an incremental loading proc-
ess is applied on a displacement-control path. It is realized by moving downward the
rigid plate with simultaneous control of displacement u (Figure 4.7). The total reaction
exerted on the plate from the soil which is equal to the applied axle force is being
measured. This procedure enables to obtain force-displacement diagram P-u applied
in procedure of the load carrying capacity assessment described in chapter 4.5.
4.5 Procedure of the load carrying capacity assessment

The procedure of the load carrying capacity assessment proposed in this report is
based on the ultimate load Pult defined as the highest value of the live load P ob-
tained in the loading process described in the previous chapter. The value is deter-
mined by the highest point at the force-displacement relation presented in Figure 4.8
(light green line). In the presented numerical examples usually several centimetres of
vertical displacement provides the desired pick value of the load. The same proce-
dure is conducted first for undamaged and then for damaged structure. Such ap-
proach enables evaluation of a defect influence on the structure load capacity as a
measurable numerical change in the maximum load carried by it.
Pult
Pdult
Figure 4.8 Axle load-displacement relation (P-u) with marked ultimate load value Pult
A significant assumption made in this report is that the changes of the ultimate load
are proportional to changes of the load carrying capacity. According to most of codes
and regulations the load carrying capacity is a equal of the ultimate load divided by a
coefficient of safety. Taking this coefficient as a constant value a ratio between the
ultimate load of damaged and undamaged structures is equal to the ratio of the cor-
responding load carrying capacities. The former relationship is proposed to be ana-
lysed directly as a measure for the influence of defects on the load carrying capacity.
Hence a defect impact factor given on a percentage basis is defined:
Pult Pultd Pult

= 100% = 100% eq. 17
Pult Pult
where Pult is a difference of the ultimate loads for undamaged Pult and damaged
Pdult structure of the same geometry and material properties.
Percentage values of are given in the diagrams presenting results of the numerical
examples of modelling defects.
For each defect various locations of the live load are analysed and the critical one,
corresponding to the minimum value of Pult, is taken into account. Therefore relying
on symmetry of the structure only defects located in a half of the arch comprise vari-
ous cases. Cases from the other half give repeating solutions.
4.6 Comparison with other methods

A comparison of the proposed method was done with popular software applied for
analysis of masonry arch bridges RING (Gilbert 2001) and Archie-M (Harvey 2001).
The analysis was carried out for a structure with the following parameters:
span: L0 = 5 m,
rise to pan ratio: r/L0 = 1/2, 1/4, 1/6,
arch barrel thickness: d = 45 cm,
height of the fill over crown: h = 35 cm,
Material parameters are the same as those given in chapter 4.4.
Result of the comparison are presented in Table 4.1 in a form of the ultimate value of
a single axle load set in a quarter point of a structure.
Table 4.1 Comparison the ultimate loads achieved by mans of various methods {kN]
r/L0 FEM RING* Archie-M
1/2 424.0 412 (406) 445
1/4 434.6 459 (398) 440
* - values in bracket for Boussinesq load dispersion
4.7 Modelling of selected types of defects

4.7.1 Deterioration
4.7.1.1 Parametric description

Deterioration is defined by means of three main parameters given in chapter 4.3 and
one additional parameter Deterioration level being a measure of material strength
value decrease:
Deterioration level () measure of material strength decrease as a ratio between
material strength value decrease fc,t to the designed value of material strength fc,t,
given on the percentage basis (eq. 18):
fcd,t
d = 100% eq. 18
fc,t
4.7.1.2 Computer modelling

Considered type of deterioration strength reduction is represented in FE model in
two manners: by decreasing values of both (proportionally) compressive fc and ten-
sile ft strength of the masonry materials and by reduction of their modules of elasticity
Eu and Ej.
It is assumed that both components of masonry i.e. masonry blocks and joints in a
given deteriorated area of a structure are together losing their strength proportionally
(Figure 4.9). This assumption seems to be reasonable taking into account fact that
usually a whole part of a structure gets deteriorated e.g. due to water penetration.
elements with
modified mate-
rial properties
Figure 4.9 Modelling deterioration
4.7.1.3 Numerical examples

Numerical examples for analysis of the deterioration influence on the load carrying
capacity are carried out for following parameters:
rise to span ratio r/L0 equal to 1/4 and 1/2,
Location in range 00,5,
Extent in range1050 %,
Intensity equal to constant value 100 %,
Deterioration level equal to constant value 70 %.
Results of the analyses are presented in Figure 4.10 for r/L0 = 1/4 and Figure 4.11 for
r/L0 = 1/2 as a value of a defect impact factor .
a)
r/L0=0.25 m, Destr=0.3
[%]
b) 17
20 16
15 15
10
14
5
13
0
12
15
14 Ext=0.3
11
13
12
10
11
10 9
L0 [m]
9
0.5
8 0.45 8
0.4
7 0.35
0.3
0.25
6 0.2 Location 7
0.15
0.1
5 0.05
0
r/L0=0.25 m, Destr=0.3
20
[%]
c) 19
20
18
15
17
10
5 16
0
15
15 Ext=0.5
14
14
13
12
13
11
10
L0 [m]
9 12
0.5
8 0.45
0.4
7 0.35 11
0.3
0.25 Location
6 0.2
0.15
0.1
5 0.05
0
Figure 4.10 Results for deterioration, r/L0 = 1/4, = 70 % a) = 10%, b) = 30 %, c)

= 50 %.
a)
r/L0=0.5 m, Destr=0.7
b) 18
[%]
20
16
15
14
10
5
12
0
15
Ext=0.3 10
14
13
12
8
11
10
L0 [m]
9
6
0.5
8 0.45
0.4
7 0.35
0.3
0.25
6 0.2 Location
0.15 4
0.1
5 0.05
0
r/L0=0.5 m, Destr=0.7
[%] c) 19
20
18
15
17
10
16
5
15
0
14
15
Ext=0.5
14
13
13
12
12
11
10 11
L0 [m]
9
0.5
8 0.45 10
0.4
7 0.35
0.3
0.25 Location
6 0.2 9
0.15
0.1
5 0.05
0
Figure 4.11 Results for deterioration, r/L0 = 1/2, = 70 %: a) = 10%, b) = 30 %, c)

= 50 %.
4.7.1.4 Conclusions from the analyses

In the analysis of arches with deterioration the influence of the defect location on the
ultimate load is similar for structures with various spans for both r/L0 = 1/4 as well as
r/L0 = 1/2. The critical location (giving the highest value of ) is dependent on the de-
fect extent. E.g. for r/L0 = 1/4 and = 10 % the critical location is equal to 0.25-0.30
but for = 30 % the critical location is equal to 0.40-0.45. This change is related to
covering by the damaged area one or two plastic hinges being formed at the failure.
In cases of = 50 % for both r/L0 = 1/4 and r/L0 = 1/2 the defect impact factor
reaches 20 %.
4.7.2 Discontinuity

Discontinuity types representing longitudinal fracture is defined by means of three
main parameters given in chapter 4.3 and one additional parameter Separation
coefficient being a measure of the arch barrel width B reduction caused by fractur-
ing:
Separation coefficient () measure of the arch barrel width B reduction caused by
separation of the edge part from the central part of the arch barrel (Figure 4.1), given
on the percentage basis (eq. 19):
B Bc B
c = 100% = 100% eq. 19
B B
where B, Bc and B are defined in Figure 4.1:
Figure 4.12 Top view of the arch barrel describing discontinuity parameters

Discontinuity represents longitudinal fracture in the arch barrel which is mostly pre-
sent near the edge of the arch barrel in the plane of the inside surface of the spandrel
wall. It is usually an effect of spandrel wall tendency to separate from the arch barrel.
Most commonly it appears symmetrically on some length of the arch near the crown
or in the whole arch barrel length.
This type of defect It is proposed to be modelled by thickness decrease for selected
2D elements representing reduction of the width of the arch barrel B on the corre-
sponding length (Figure 4.13). Reduced width is equal to the ratio between effective
width of the arch barrel arisen by fracturing Bc to the designed width B (eq. 19) multi-
plied by 1 meter. In this way it is assumed that there is no more interaction between
separated edge and the remaining central part of the arch barrel.
elements
with reduced
thickness
Figure 4.13 Modelling discontinuity

This approach assumes that both components of masonry arch i.e. masonry blocks
and joints are fractured together in a given plane of a structure what often represents
real cases.

Numerical examples for analysis of the discontinuity influence on the load carrying
capacity are carried out for following parameters:
Location equal to constant value 0,5,
Extent in range 070 %,
Intensity equal to constant value 100 %,
Separation coefficient equal to constant value 20 %.
r/L0=0.25 m, Separ. coeff.=0.2
[%] 14
15
12
10
10
8
0
15
14
6
13
12
11 4
10
L0 [m]
9
0.7
8 2
0.6
0.5
7
0.4
6 0.3
0.2
Extent
5 0.1
0
Figure 4.14 Results for discontinuity, r/L0 = 1/4, S = 20 %
r/L0=0.5 m, Separ. coeff.=0.2 16
[%]
14
15
12
10
5 10
0
8
15
14
13 6
12
11
4
10
L0 [m]
9
0.7
8
0.6 2
7 0.5
0.4
6 0.3
0.2
Extent
5 0.1
0
Figure 4.15 Results for discontinuity, r/L0 = 1/2, S = 20 %


From the results a main conclusion can be drawn that the influence of discontinuity
on the ultimate load is more significant for structures with longer spans in case of
arches with r/L0 = 1/4 as well as with r/L0 = 1/2. For spans equal to 15 m the defect
impact factor exceeds 15 %.
In all the analysed cases increase of the defect extent for a given structure is accom-
panied by increase of the defect impact factor value what follows predictions.
4.7.3 Loss of material

Losses represent loss of masonry blocks and joints. It is described only by means of
three main parameters given in chapter 4.3.

Considered type of losses concerns a whole part of masonry (including blocks and
joints). Only loss of material from intrados side is analysed what reflects real cases.
This type of a defect is modelled by removing of finite elements in the arch barrel
height in proper areas involving both masonry blocks and joints (Figure 4.16). It is
important not to change shape nor size of remaining elements which are always
slightly influencing the solution. Therefore a mesh of intact structure should take into
account localisation and size of losses analysed in next step.
removed
elements
Figure 4.16 Modelling losses

Numerical examples for analysis of the losses influence on the load carrying capacity
are carried out for following parameters:
Location in range 0,050,50,
Extent equal to constant value 10 %,
Intensity equal to 10 % and 20 %.
a)
b)
Figure 4.17 Results for loss of material, r/L0 = 1/4, = 10 % a) = 10%, b) = 20 %

a)
b)
Figure 4.18 Results for loss of material, r/L0 = 1/2, = 10 % a) = 10%, b) = 20 %


In the analyses of arches with loss of material the influence of the defect location on
the ultimate load is very similar for structures with various spans for both r/L0 = 1/4 as
well as r/L0 = 1/2. The critical location (giving the highest value of ) for r/L0 = 1/4 is in
range = 0.25-0.50 and for r/L0 = 1/2 is in range = 0.35-0.50. The influence of the
location on the limit state can be explained by analysis of a plastic hinge location
against the damaged area what is presented in Figure 4.19 for three cases of =
0.125, 0.250 and 0.500 for a bridge with r/L0 = 1/4. From the comparison of the re-
sults for in Figure 4.17 and modes of failure presented in Figure 4.19 it can be
noted that the highest influence of the defect takes place when a plastic hinge ap-
pears at the location of the defect. For location of the defect = 0.125 plastic hinges
appear out of the damaged area, therefore value is equal almost to zero. For loca-
tion of the defect in the whole range = 0.250-0.500 a plastic hinge arises exactly
within the damaged area and therefore the defect impact factor gets high values.
= 0.125 a)
b)
= 0.250
c)
= 0.500
Figure 4.19 Results for loss of material, r/L0 = 1/4, a) = 0.125, b) = 0.250,
c) = 0.500
For r/L0 = 1/4 and = 0.40 (with = 10 %, = 20 %) the defect impact factor reaches
30 % and for r/L0 = 1/2 and = 0.50 the defect impact factor reaches 25 %.
5 Final remarks
5.1 Conclusions
Commonly available methods for the assessment of masonry bridges provide possi-
bilities of simplified modelling of the main defect types. Selection of the appropriate
method is strongly dependent on the considered defect type. Some of the simple
methods are especially useful in the analysis of the typical defects providing suffi-
ciently precise results for comparison of the various defect cases. However some of
the defects require more advanced approach like those based of 3D FE models to
evaluate reliably their influence on the load carrying capacity of the structure.
Detailed analysis of the influence on the ultimate load of the selected types of defects
with their the most common parameter values is presented. It shows that such typical
defects can reduce the ultimate load of the structure by up to 30 % what is a signifi-
cant value indicating the need for consideration of this phenomenon. Also an impor-
tant finding is the most crucial locations for the defects. For all the analysed defect
types and structure geometries the critical locations are equal usually to about =
0.40 which coincides with location of one of the forming plastic hinges.
5.2 Plans for further works

For further analyses more extensive parametric study of the defects influence on the
ultimate load is recommended. It should be extended to larger number of the defect
parameter values and more diverse bridge geometries.
The influence of other defect types like the arch barrel deformation and supports
displacements could be studied by means of FEM including 3D models.
References
Ali S., Page A. W. (1988): Finite element model for masonry subjected to concen-
trated loads. Journal of Structural Engineering, 114, No. 8, pp.1761-1784.
Azevedo J., Sincraian G. (2001): Modelling The Seismic Behaviour Of Monumental
Masonry Structures. International Millennium Congress - Archi 2000, Paris.
Brencich A., De Francesco U. (2004a): Assessment of Multi-Span Masonry Arch
Bridges. Part I: a Simplified Approach, J. of Bridge Eng.ng, ASCE, 9, pp. 582-
590.
Brencich A., De Francesco U. (2004b): Assessment of Multi-Span Masonry Arch
Bridges. Part II: Examples and Applications. J. of Bridge Eng.ng, ASCE, 9, pp.
591-598.
Brencich A., De Francesco U., Gambarotta L. (2001): Non linear elasto-plastic col-
lapse analysis of multi-span masonry arch bridges. Arch'01, Paris, 19-21 Sep-
tember, pp. 513-522.
Choo, B.S., Coutie, M.G. and Gong, N.G. (1991a): Finite element analysis of ma-
sonry arch bridges using tapered beam elements. Proc. Inst. Civ. Engs. Part 2,
91, pp. 755-770.
Choo, B.S., Coutie, M.G. and Gong, N.G. (1991b): The effects of cracks on the be-
haviour of masonry arches. Proc. 9th International Brick/Block Masonry Confer-
ence, Vol. 2, pp. 948-955, Berlin.
Crisfield M. A. (1985): Finite element and mechanism methods for the analysis of
masonry and brickwork arches. Transport and Road Research Laboratory, Lon-
don.
Department of Transport (2001): The Assessment of Highway Bridges and Struc-
tures, Design Manual for Roads and Bridges: BA 16/97. Volume 3. Section 4.
Part 4.
Fanning, P.J., Boothby, T.E. (2001): Three-dimensional modelling and full-scale test-
ing of stone arch bridges. Computers & Structures, Vol. 79, pp. 2645-2662.
Fanning, P.J., Boothby, T.E., Roberts, B.J. (2001): Longitudinal and transverse ef-
fects in masonry arch assessment. Construction and Building Materials, Vol. 15,
pp. 51-60.
Ford T.E., Augarde C.E., Tuxford S.S. (2003): Modelling masonry arch bridges using
commercial finite element software. 9th International Conference on Civil and
Structural Engineering Computing, Egmond aan Zee, The Netherlands.
Frunzio G., Monaco M., Gesualdo A. (2001): 3D F.E.M. analysis of a Roman arch
bridge. Historical Constructions, P.B. Loureno, P. Roca (Eds.), Guimares, pp.
591-599.
Gilbert, M. (1998): On the analysis of multi-ring brickwork arch bridges. Proc. 2nd
International Arch Bridges Conference, Venice, pp. 109-118.
Gilbert, M. (2001): RING home page, http://www.shef.ac.uk/ring
Gilbert, M., Melbourne, C. (1994): Rigid-block analysis of masonry structures. The

Structural Engineer, 72, 356-361.
Harvey W. (2001): Obvis home page, http://www.obvis.com
Heyman, J. (1982): The Masonry Arch. Ellis Horwood.
Hibbit, Karlsson, Sorensen (2005): ABAQUS 6.5 Documentation. Hibbit, Karlsson &
Sorensen Inc.
Hillerborg A., Modeer M., Petersson P. E. (1976): Analysis of Crack Formation and
Crack Growth in Concrete by Means of Fracture Mechanics and Finite Ele-
ments. Cement and Concrete Research, Vol. 6, pp. 773-782.
Lee J. S., Pande G. N., Middleton J., Kralj B. (1996): Numerical modelling of brick
masonry panels subject to lateral loadings, Comp. Struct., 61, pp. 735-745.
Livesley R.K. (1978): Limit analysis of structures formed from rigid blocks. Int. Journ.
for Num. Meth. in Eng., 12, pp. 1853-1871.
Livesley, R.K. (1992): A computational model for the limit analysis of three-
dimensional masonry structures. Meccanica 27, pp. 161172.
Loo, Y. and Yang, Y. (1991): Cracking and failure analysis of masonry arch bridges.
ASCE J. Struct. Eng., 117, No. 6, pp. 1641-1659.
Lourenco P.B., Rots J.G., Blaauwendraad J. (1995): Two approaches for the analysis
of masonry structures: Micro and macro-modelling. HERON 40(4): 313-340.
Markeset G., Hillerborg A. (1995): Softening of concrete in compression - localization
and size effects. Cement and Concrete Research, Vol. 25, No. 4, pp. 702-708.
Martn-Caro J. A., Martnez J. L., Len J. (2004): A First Level Structural Analysys
Tool for The Spanish Railways Masonry Arch Bridges. Proc. 4nd International
Arch Bridges Conference, P. Roca and E. Oate (Eds), CIMNE, Barcelona.
Martn-Caro J. A., Martnez J. L., Len J., Lopez L. (2002): A Methodology for the
Assessment of Masonry Arch Bridges.
Ng K. H., Fairfield C. A., Sibbald A. (1999): Finite-element analysis of masonry arch
bridges. Proc. Inst. Civ. Engrs Structures and Buildings 134, pp. 119-127.
O'Dwyer, D. (1999): Funicular analysis of masonry vaults. Computers and Structures,
Vol. 73, pp. 187-197.
Orduna A., Lourenco P. (2005a): Three-dimensional limit analysis of rigid blocks as-
semblages. Part I: Torsion failure on frictional interfaces and limit analysis for-
mulation. International Journal of Solids and Structures 42, pp. 51405160.
Orduna A., Lourenco P. (2005b): Three-dimensional limit analysis of rigid blocks as-
semblages. Part II: Load-path following solution procedure and validation. Inter-
national Journal of Solids and Structures 42, pp. 51615180.
SB-CAI (2007): Guideline for Condition Assessment and Inspection. NDT-toolbox for
the Inspection of Railway Bridges. Deliverable 3.16. Prepared by Sustainable
Bridges a project within EU FP6, Available from: www.sustainablebridges.net.
Thavalingam A., Bicanic N., Robinson J. I., Ponniah D. A. (2001): Computational
framework for discontinuous modelling of masonry arch bridges. Computers
and Structures, 79, pp. 1821-1830.

UIC (1994): Code 778-3R. Recommendations for the assessment of the load carry-
ing capacity of existing masonry and mass-concrete arch bridges.
Van der Pluijm R. (1997): Non-linear Behaviour of Masonry under Tension. HERON,
Vol. 42, No. 1, pp. 25-54.
Potentiality of Probabilistic Methods in the
Assessment of Masonry Arches
PRIORITY 6
INTEGRATED PROJECT

Author/s: J. R. Casas, P. Roca, C. Molins, UPC
Revision date:

Dissemination Level
PU Public X
Table of Contents
1 Introduction ..........................................................................................................................4
2 Objectives ............................................................................................................................5
3 Failure modes and Limit State Functions ............................................................................6
3.1 Single ring arches.......................................................................................................6
3.2 Multi-ring arches.........................................................................................................7
3.2.1 Four hinge mechanism...................................................................................8
3.2.2 Ring separation ..............................................................................................8
4 Levels of assessment: failure criteria and theoretical models .............................................9
4.1 Axial- Bending failure .................................................................................................9
4.1.1 Monotonic loading ..........................................................................................9
4.1.2 Cyclic loading (fatigue) .................................................................................13
4.2 Ring separation ........................................................................................................13
4.3 Theoretical models ...................................................................................................14
4.3.1 Probabilistic approach ..................................................................................15
4.3.2 Simplified numerical models.........................................................................17
5 Basic variables...................................................................................................................54
5.1 Dimensions...............................................................................................................54
5.2 Materials...................................................................................................................54
5.3 S-N diagrams. Probabilistic fatigue models..............................................................54
5.3.1 S-N diagrams for masonry in compression-bending ....................................54
5.3.2 S-N diagrams for masonry in shear..............................................................65
6 Examples ...........................................................................................................................68
6.1 Assessment of ULS by 4-hinges mechanism ...........................................................68
6.1.1 Description of the bridge ..............................................................................68
6.1.2 Assessment assuming linear-elastic behaviour ...........................................70
6.1.3 Assessment based on non-linear response and system failure analysis .....71
6.2 Assessment of Fatigue LS .......................................................................................75
7 Conclusions .......................................................................................................................78
8 References ........................................................................................................................79
1 Introduction
Several methods are available for the assessment of capacity of masonry arch bridges, from
very simple to more sophisticated, as presented in [1]. In this document, only the methods
based on the use of probabilistic procedures are presented to investigate their potentiality of
application in the assessment of masonry arches.
Techniques using reliability-based methods have been successfully applied to the assess-
ment of concrete and steel bridges. Nowadays, a lot of examples present in the available
technical literature show the advantages of the use of reliability-based approaches to the
assessment of existing bridges. However their application to the masonry bridges has been
almost negligible. One of the reasons is the difficulty to define reliable failure criteria for this
type of structures. Another reason is the lack of statistical data on material properties of ma-
sonry and filling material. The experimental tests show that normally the failure of an arch is
of a global nature more than due to the failure of a bridge component. In most cases, even
the division of the bridge in different components is almost impossible. For this reason, also
the lack of accurate theoretical models for the idealization of the behaviour of the masonry
arch bridge as a system, including the interaction effects with the filling material, spandrel
walls, etc., has been another difficulty not yet overcome. Last, but not least, the absence of
reliable data on the statistical definition of the material properties has been another issue that
has limited the use of probability-based assessment techniques. Very few data is available
for the behaviour of the masonry under static loads. The lack of experimental data is even
more dramatic in the case of cyclic loading.
However, some experiences have shown the potentiality of the use of reliability-based as-
sessment together with non-linear models of the bridge behaviour in the capacity assess-
ment of masonry arch bridges. In some cases, masonry bridges that will be rated as unsafe
when using standard methods of assessment and linear models, have reported high enough
level of safety when appropriate theoretical models and probabilistic assessment tools are
applied [2,3,4]. The global failure of the arch due to the formation of a hinged mechanism has
been the only failure mode considered so far. However, other response mechanisms (ring
separation, material crushing,) may lead the masonry arch to the failure and will be also
investigated in the present work. Finally, the serviceability issues related to the fatigue be-
haviour under service loads will be also considered, as they appear also to play an important
role in the service life of existing masonry arches [1]. As a consequence, a probabilistic ap-
proach to the fatigue failure assessment of masonry arch bridges is also presented in this
document.
2 Objectives
Based on the existing achievements obtained till now in the use of reliability-based methods
for the assessment of existing bridges, the main objective is to investigate the potentiality of
probabilistic methods in the assessment of masonry arches, leading to an increase of the
load-capacity of these structures by comparison with the results from deterministic analyses.
To this end, the following objectives are considered:
1.- Definition of failure modes and limit state functions for masonry arch bridges
2.- Definition of mechanical models for accurate simulation of bridge behaviour up-to failure,
according to the identified failure modes in 1
3.- Derivation of simplified models for the reliability-based assessment of masonry arch
bridges
4.- Statistical definition of resistance random variables involved in the limit state functions.
3 Failure modes and Limit State Functions

The whole failure mechanisms that we may find in masonry arch bridges are presented else-
where [D4.7.1]. It is usually assumed for a single square spanned bridge that the failure
mechanism involves the formation of 4 hinges. However, there are other failure mechanisms
as the formation of 3 hinges plus the horizontal displacement of the abutment. Other ways of
failure include the crushing of the masonry, the ring separation in multi-ring arches due to the
failure of the mortar between rings, punching shear failure as a result of failure of the radial
mortar joints, foundation failure because of scour, failure of the backfill and others. In sum-
mary, the various modes of failure most reported are the four-hinge mechanism, the ring
separation in multi-ring arches and the slippage at the foundations. Because of the lack of
reliable material data (in the statistic sense) or available response mechanisms, here, only
those more prone to be analyzed using reliability-based methods are shown: four-hinge
mechanism and ring separation. Basically, the most likely failure mode will depend on the
type of construction used for the arch. For this reason, a separation is made between single
and multi-ring arches. Also, a different failure mode could appear in the case of static or cy-
clic loading.
3.1 Single ring arches

In the case of an arch built with an unique ring, either made of a single masonry ( see figure
1) or of several interrelated units of masonry ( see figure 2), the most likely failure is due to
the formation of a four-hinge mechanism because of an extreme load, as presented in figure
6. The mechanism normally appears before any separation between units or crushing of the
material occurs. This type of failure should be considered as an Ultimate Limit State (ULS).
However, the degradation and loss of structural integrity due to ageing, may affect the ability
of the bridge to carry the normal service loads for the expected life of the bridge. In this case,
the resulting limit state is considered as a Serviceability Limit State (SLS) or also a Permissi-
ble Limit State (PLS), in the sense that even not reaching the maximum allowable load for a
unique application, the bridge may fail due to repeated application of lower levels of loading.
In this case, the attempt is to determine the residual life taking into account the normal loads
in the bridge and looking to its long-term performance [21]
Figure 1.- Single ring arch formed by an unique row of units

Figure 2.- Single ring arch formed by several units across the arch barrel depth
Despite being less feasible, the fatigue failure has to be also considered in this type of
arches, mainly when deterioration due to aging and service loads may derive on lower fa-
tigue strength of the masonry mix. The fatigue failure is due to repeated cycles of compres-
sion causing the crushing and squeezing of the mortar from the mortar joints followed by
vertical splitting of the bricks.
3.2 Multi-ring arches

In the case of a multi-ring arch (figure 3), the failure could be either by the formation of the
four-hinge mechanism or by ring separation. The failure by one or other mechanism will de-
pend on the specific characteristics of the bridge: span-length, type of mortar between rings,
and also the loading type: static or dynamic.
Figure 3.- Multi-ring masonry arch bridge

3.2.1 Four hinge mechanism

The failure mechanism is similar to the case of a single arch (see figure 4) and again the
failure should be analyzed as an ULS. In a multi-ring masonry arch, normally this mechanism
appears in the case of monotonic loading conditions.
Figure 4.- 4 hinges mechanism of failure of a multi-ring arch (Photo: USAL)
3.2.2 Ring separation

This failure mechanism (see figure 5) normally appears under cyclic loading conditions, al-
though also a monotonic extreme load may cause the separation. In the case of static load-
ing, ring separation can occur between the point and the nearest abutment and the failure
is considered as an Ultimate Limit State. For cyclic loading, the separation occurs between
the and point. In the case of the ring separation under cyclic loading, the failure should
be considered a Fatigue Limit State, or, again, as a Permissible Limit State.
Ring separation
Figure 5.- Failure by ring separation in a multi-ring arch ( Photo USAL)

4 Levels of assessment: failure criteria and theoretical

models
Depending on the criteria adopted to define the failure of the bridge, different limit state func-
tions and different levels of assessment can be formulated. The process follows the step by
step model as presented in chapter 3 of the Guideline. Increasing levels of assessment de-
rive in more accurate results, i.e. in bridge response closer to reality. Each level of assess-
ment requires a corresponding theoretical model. According to the failure mechanisms de-
fined in 3, the possible levels of assessment are described below.
4.1 Axial- Bending failure
4.1.1 Monotonic loading

This failure occurs when an extreme load causes excessive bending moments/shear forces
combined with important axial loads resulting in the failure either by formation of a mecha-
nism or excessive compression in the material, provoking crushing of the masonry.
The failure criteria will be based on the Ultimate Limit State (ULS) formulation. At least, 3
assessment levels exist, each one with increasing level of difficulty and accuracy depending
on the failure criteria and analytical model used.
Linear at cross- section level

The lowest capacity level will be obtained considering the failure of the bridge when any of
the fibres of any cross-section is in tension. The values of the solicitations (bending mo-
ment, normal force) and the distribution of stresses in the cross-section are derived assum-
ing a linear behaviour of the structure and the material. Of course, this is an extremely re-
strictive and over-conservative failure criteria. However, if the bridge passes this level of as-
sessment, that requires very simple modelling tools, it may be assured that the bridge is in
very good shape.
Non-linear at cross-section level

If the bridge does not pass the first level of assessment, then the next level is applied. In this
case, the failure mechanism is defined to occur when the maximum compression stress ex-
ceeds the maximum allowable stress in the material, or alternatively, when the compression
block in the masonry is not able to balance the normal load (N) and bending moment (M) due
to the external actions. Failure is reached when any cross-section fails. The internal forces N
and M are obtained with a linear-elastic model of the arch. The Limit State Function can be
formulated in the following way:
N 2M
G=H (1)
fc B N
H is the cross-section depth, B is the cross-section width, fc is the compressive strength of

masonry. Provided that a statistical definition for all random variables involved is available,
the reliability index can be evaluated by FORM (First Order Reliability Method). In chapter
6.1.2 is shown a practical application of this method.
Non-linear at system level

In this case, and if the ultimate strength of the material is not reached at any section, the
bridge is considered to fail when a sufficient number of plastic hinges (normally 4) develop in
different cross-sections to form a global/system failure mechanism (see figure 6). The failure
can be due to either the development of a mechanism or the incapacity to accommodate
further deformations without breaking the material.
Figure 6.- Collapse of the Prestwood bridge due to 4 hinges mechanism [5]
To take into consideration the behaviour up to failure of the bridge, more sophisticated theo-
retical models are necessary that take into account both material and geometrical non-
linearities. These models are able to simulate the formation of the hinges in the critical sec-
tions, as seen in figures 7 and 8 [4]. The bridge in figure 7 is a multi-span masonry-brick arch
bridge and the bridge in figure 8 is a single-span masonry- stone arch bridge. In both cases
the load is applied at the quarter span.
In figure 7, the black zones affecting the overall cross-section depth indicate the formation of
a hinge.
(a)
b)
Figure 7.- Magarola arch bridge (Spain). General view (a) and stresses at failure (b).
(a)
(b)
Figure 8.- Jerte arch bridge (Spain). General aspect (a) and stresses at failure (b)
In figure 8, the hinges are located in the cross-section with a full cross-section depth col-
oured in red.
The application of probability-based assessment methods when using such sophisticated
and advanced models is cumbersome. Because of the non-linear nature of the problem, an
explicit limit state equation is not available. The wisest approach is to use a simulation proc-
ess for the calculation of the probability of failure. However, crude Monte-Carlo simulation
can be hardly applied due to the huge number of runs necessary in the non-linear model and
the complexity of such a model. For this reason more advanced simulation techniques as the
importance-sampling, directional sampling and Latin Hypercube Method or other alternative
methods as the Surface Response method [6] are the most adequate tools to combine with a
full non-linear model. In chapter 6.1.3 is presented a practical application using the Latin Hy-
percube Method.
Simplified models to assess the ultimate capacity taking into account the non-linear behav-
iour are available through their calibration from the results of a full rigid block and plastic
analysis. This is the case of the model developed by Martn-Caro et al. 2004 [7] . Explicit
equations are proposed for the maximum point and uniform distributed loads in terms of sim-
ple geometrical parameters and material strength. The model is of application in the case of
not skewed and not curved bridges, not multi-ring masonry bridges with consecutive regular
spans. The abutments should be un-deformable and the spandrels are solid. The parametric
study comprises spans in the range from 2 to 20 m., with a minimum rise to span ration of
1/6 and vault depth at crown to span ratio limited to a value between 0.05 and 0.10 depend-
ing on the span-length. The spandrel high at crown is in the range from 0.15 to 2 m. and the
maximum pier height is 10 m. Despite looking very restrictive, these requirements are ac-
complished by an important number of masonry bridges in the network.
The use of simplified models allows in some cases an explicit equation for the bridge capac-
ity, as a function of the geometric and material characteristics. In this way the load and resis-
tance parts of the problem become uncoupled and a Limit State Function in the form G=R-S
can be formulated. The reliability index can be then evaluated using FOSM or FORM.
A complete summary of the available numerical and analytical models to analyze the four-
hinge failure mechanism is reported in [8].
4.1.2 Cyclic loading (fatigue)

The few available high cycle fatigue tests either in small specimens [9] or full laboratory
models [1] have shown that the fatigue strength of brick masonry subject to compressive-
bending state depends upon the induced stress range, the mean or maximum induced stress
and the quasi-static compressive strength of the masonry under similar loading conditions.
4.2 Ring separation

Ring separation in multi-ring masonry arches may occur either by an increasing monotonic
load (ULS) or by the repetition of a great number of small cyclic loads (Fatigue LS). In both
cases the failure is due to the rupture of the mortar between the rings due to the shear
stresses at this part of the bridge. From the point of view of the analysis, two cases are of
consideration as has been reported in several tests [10]:
1.- The arch fails by ring separation independently of the number of cycles of load.
2.- The arch may develop the two failure mechanisms. For a reduced number of cycles of
load, the failure is by four-hinge mechanism with a load level close to the static failure load,
and for a high number of cycles with lower level load, failure happens due to ring separation.
In the first case, only one failure mode is of consideration and the limit state of fatigue can be
used to predict the bridge capacity. The use of S-N curves similar to the case of steel struc-
tures under fatigue may be of application [10]. The main issue is to obtain the S-N curves,
which implies an important testing work. From the few tests in small scale arches reported so
far it seems than the S-N curve has only one branch. This is different from the case of steel
where several branches with different slopes have been reported (mainly for prestressing
and reinforceing steels [11]. The case of several number of cycles of load with different level
could be approached by using the Miners rule as only one mechanism of damage deriving in
the ring separation is present in this case.
In the second case, two failure modes are coupled. In this case, although a S-N diagram
could be still derived, with two branches, each one modelling one failure mode (see figure 9)
the use of Miners rule could be debatable. In fact, the accumulation of damage is not of ap-
plication if damage is caused by different mechanisms. Only the damage caused by the
same mechanism of failure can be accumulated. On the other hand, one external load pro-
duces some damage in both mechanisms. In this case, the S-N diagrams may have one of
the two shapes shown in figure 9. For a low number of cycles with very high load, the arch
fails by the 4-hinge mechanism, up to a certain number of cycles of load (N*) where the criti-
cal mechanism becomes the ring separation.
Figure 9.- S-N curves for the coupled mechanism-ring separation fatigue behaviour
However, we should mention that this approach based on the definition of S-N curves for the
whole arch derives on an inapplicable methodology. In fact, the shape of the S-N curve in the
case of the bridge (system level) will depend on the specific characteristics of the bridge, with
its particular configuration of geometry, boundary conditions and loading pattern. Depending
on those parameters (among others), the fatigue effects due to the hinged mechanism or the
ring separation will be more or less enhanced. In fact, the level of the normal or shear
stresses governing both mechanisms depend on the mentioned variables. As a conse-
quence, the S-N curve obtained in a fatigue test will be only representative of the specific
conditions of such test, and will be hardly adopted in the case of a bridge with other condi-
tions. For this reason, the approach based on the definition of S-N curves at a specimen or
material level and not at the arch level will be adopted as presented in the next chapter. In
this case, the derivation of the S-N curves become also simpler as the fatigue tests may be
carried in small specimens and not in the whole arch structure.
4.3 Theoretical models

The system level model of analysis, as the one presented in chapter 4.1 for the failure analy-
sis at the ULS due to the formation of an hinged mechanism, is of application only for this
particular failure mode of the arch because the failure mode itself requires the participation of
the arch as a whole. However, other failure mechanisms are possible that may involve failure
mechanism at a more local level. In fact, the following failure mechanisms are possible in the
arch:
- Hinged mechanism
- Ring separation
- Crushing of masonry
- Sliding of masonry blocks
From them, the first one should be analysed as an ULS as described in 4.1, but the other
three can be analysed from a material point of view involving a Fatigue or Serviceability LS at
service loads, avoiding in this way the problems deriving from the use of S-N curves at the
system level as mentioned in chapter 4.2
In fact, ring separation and sliding of masonry is provoked by the failure of the shear strength
in the interface between masonry and mortar. Crushing is controlled by the compressive
strength of masonry. In both cases, failure may appear due to a repeated number of cycles
of relatively low level of load. Depending on the specific characteristics of the bridge (geome-
try, boundary conditions, loading,.) the variable external loads will produce higher or lower
levels of compression and shear stresses in the material. The level of stress in each case
and the corresponding number of load cycles jointly with the fatigue strength will finally de-
cide if the first mechanism of failure to appear is the crushing of masonry or the sliding or ring
separation.
Therefore, the following methodology is proposed to model the behaviour of the failure
mechanism of ring separation, sliding and crushing when they are caused by accumulation of
cycles of load at service level (fatigue effect):
1.- Divide the problem in two analysis: one related to compression and the other to shear
interface stress in the masonry
2.- Use the Miners rule ( accumulated damage) as criteria for failure for both the compres-
sive strength and shear strength of the masonry
3.- Obtain S-N curves for both mechanisms of fatigue failure ( compression and shear)
4.- Identify in the bridge the more prone locations to suffer from excessive compression and
shear stresses
5.- Using a numerical model of the bridge, obtain the number of cycles and the stress incre-
ment (for both compression and shear) in the identified locations due to the external variable
loads. The rain-flow method can be used to this end.
6.- The bridge will fail when either the accumulated damage of Miner in compression or in
shear will reach a value equal to 1.0
4.3.1 Probabilistic approach

Considering that a set of experimental points in the S-N plane is provided for the case of
compression and shear fatigue strength tests of the masonry, then the fatigue capacity from
a probabilistic point of view can be analysed in a similar way as in the case of steel [11]. The
limit state function in the case of several number of cycles of different load intensity, assum-
ing that Miners rule is of application, can be written in the following way:
ni
G= 1 (2)
i Ni
ni = number of cycles of load level i due to external loads (random variable)
Ni = number of cycles of load level i that the bridge can support (random variable)
ni as a random variable will be obtained via structural analysis taking into account the ran-
domness in the live-loads acting on the bridge or, alternatively, by extrapolation of measure-
ments taken in the bridge. Also the variability of the bridge properties should be considered
in the definition of ni as the resultant stress increment will also depend on this properties.
The way to derive the statistics of this random variable is by simulation of traffic effects jointly
with simulation of geometric and material properties of the structure [11,12]. In the case of
railway traffic and taking into account that the stress increments due to bridge dynamics will
be very low for this type of bridges ( the infill mitigates the vibration level), one may consider
the passage of each convoy as a cycle of loading.
The process of simulation to derive the random variable ni requires the use of a theoretical
model where the input variables (geometry and mechanical properties of masonry) can be
modified according to their randomness. Several advanced FEM Codes are available for the
analysis of masonry arches. However in order to make simulation of practical application, a
numerical model as simple as possible is essential. The possibility of the use of simplified
models of analysis of masonry arches is discussed in chapter 4.3.2
The statistical definition of Ni can be done in the following way. Different works have shown
that the Weibull distribution function agrees very well with the expected physical criteria of
progressive fatigue deterioration [13,14]. On the basis of physically valid assumption, sound
experimental verification, relative ease in its use and better developed statistics, the Weibull
distribution has been widely used for the fatigue analysis of metals. It is also well-suited for
certain procedures of statistical extrapolation of large systems [13]. In [14] and [15], the
distribution of fatigue life of concrete was found also to approximately follow the Weibull
distribution. Some theoretical and experimental works [16,17] have shown also the feasibility
of the Weibull distribution regarding the statistical model for steel wires and strands to
fatigue. The randomness of the masonry' fatigue strength is also represented here by means
of a Weibull distribution. The reason for this assumption is fully demonstrated in chapter 5.3
The expressions for the Probability Density Function (PDF), fN(n), and the Cumulative
Distribution Function (CDF), FN(n), of a Weibull random variable are:
1
n n0 nn
f N ( n) = exp 0
, n n0 (3)
u n0 u n0 u n 0

nn
FN (n) =1 exp 0
, n n0 (4)
u n0
being = shape parameter, u = characteristic extreme value and n0 = minimum value. For
fatigue analysis, n0 is usually taken equal to 0.0 [13,14].
The first two moments of the variable are:
1
E ( N ) =u1 + (5)

2 1
var( N ) = u 2 1 + 2 1 + (6)

is the gamma function. For a complete definition of the fatigue strength in probabilistic
terms, the parameters should be known for all possible stress levels. This is not the case in
most situations because of lack of experimental data. In chapter 5.3, is explained how a
Weibull distribution for the fatigue strength of masonry and for several stress levels can be
obtained based on the few experimental data available at the present time.
4.3.2 Simplified numerical models

Advanced and complex Finite Element models are necessary to correctly model the re-
sponse of masonry arch bridges affected by different parameters. As an example, the con-
sideration of the infill material in the global response of the bridge, the skewness and the
modelisation of multi-ring arches requires the use of sophisticated models. However, for a
probabilistic approach of the bridge safety, involving a simulation procedure, the availability
of more simple numerical models is essential. To this end, in the present chapter are pre-
sented the works developed to calibrate simple numerical models for the analysis of masonry
arches that take into account the infill material, the presence of multi-ring arch barrels and
the bridge skewness. The calibration has been carried out from the results derived with the
advanced software DIANA.
Influence of the infill material

The work presented here has two main parts. In the first one the bridge tested at the Univer-
sity of Salford to investigate the soil-structure interaction [1] is modeled with an advanced
software for masonry analysis (DIANA). The model takes into account the soil-structure in-
teraction and the results from the test are used to calibrate the model parameters, mainly the
stiffness of the infill material. With the same model, the influence of the infill stiffness on the
horizontal distribution of pressure over the arch barrel and the ultimate load is shown. In the
second part, the results of the advanced theoretical model are used to calibrate a much sim-
pler model based on the plastic analysis theory and programmed in a excel sheet. It is shown
how the introduction in this simplistic model of the lateral forces over the arch due to the infill
action derives in results very close to the experimental ones.
Description of the experimental model

The experimental data available and used here is that obtained in the experimental pro-
gramme carried out at Salford University and fully reported in [1]. The main characteristics
are as follows:
An arch filled with limestone and one arch filled with clay were tested. The geometry is pre-
sented in figures 10 and 11.
north abutment
south abutment
1045
8300
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
plan view
215
305
1665
750
395
550 180 3373 550
730 322 3000 325 3923
8300
elevation
Figure 10.- Bridge filled with limestone (dimensions in mm)

north abutment
south abutment
8300
1045
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
plan view
215
305
2165
1665
750
395
1930 322 3000 325 2723
8300
elevation
Figure 11.- Bridge filled with clay (dimensions in mm )
Class A Engineering bricks described as Nori bricks were used in the construction of the
arch barrel. The bricks had the following mean properties: compressive strength 154 N/mm2
and density 2370 kg/m3 and the nominal dimensions 215x102x65 mm. The mortar 1:2:9
(cement:lime:sand) mix by volume was used throughout the arch barrel. The mean compres-
sive strength was 1.9 N/mm2 as determined from five 100 mm cubes cured under the same
conditions as the arch barrel.
The first test bridge was filled with MOT Type 1 graded crushed limestone. The limestone
was compacted in 11 layers and the required weight of fill for each layer was compacted to
the required specification to control its density. An average density of 1.91 Mg/m3 was
achieved.
The second test bridge was filled with clay up to the level of the crown and compacted in 100
mm layers. The clay was described as a firm reddish brown slightly sandy CLAY with occa-
sional gravel with an optimum moisture content of 9%, liquid limit 29% and plastic limit 12%.
During filling, small samples were taken at regular intervals for moisture content testing. The
average moisture content was 13.4% based on 95 samples (standard deviation 1.4%); the
average density was found to be 2.21 Mg/m3 based on 14 tube samples taken from locations
across the clay body after the test (standard deviation 0.3 Mg/m3). The filling was finished
with two layers of limestone on the top of the clay.
The load was positioned vertically above the quarter point of the arch barrels
In order to study the influence of abutment fixity, two reinforced concrete abutments on which
the arch barrel were built were fixed 3000 mm apart, parallel to each other. Each abutment
comprised two parts; a lower section that was bolted to the structural strong floor, and an
upper section that could slide should the forces be large enough to overcome the
bond/friction between the blocks. The details of the two abutments are shown in Figure 12.
154
230
215
13
530
530
20
395
322 325
south abutment details north abutment details

(width 999mm) (width 999mm)
Figure 12.- Construction details of abutments in limestone and clay filled bridge tests
Both bridges ultimately collapsed in four-hinged mechanisms (see figure 13). Whilst at col-
lapse hinged mechanisms formed, some abutment movement was recorded prior to this
stage. In the case of the crushed limestone filled bridge spreading of the abutment remote
from the applied load peaked at 3 mm, whilst in the case of the clay filled bridge this was
significantly greater, peaking at 8 mm. However, in the case of the clay filled bridge rotation
about the base of the lower abutment block remote from the applied load was identified, de-
spite the fact that this had been bolted to the structural strong floor. In the case of each
bridge, movement of the abutment closest to the applied loads was small < 1 mm.
Figure 13.- Collapse mechanism of arch barrel

Figure 14 shows the quarter span vertical deflections of both bridges tested. A peak load of
125 kN for the limestone filled arch and 85 kN for the clay filled, were obtained.
140
Bridge 1 (limestone filled)

120
100
Bridge 2 (clay filled)
Applied load (kN)
80
60
40
20
0
0 5 10 15 20 25 30
Radial displacement (mm)
Figure 14.- Load- vertical displacement (quarter span) relationship for both tests
Description of the numerical model

A 2D model of the arch barrel and the infill material was done using the commercial software
DIANA (figure 15)
Figure 15.- 2-D FEM model

The model consists of 8000 triangular elements and 200 special elements to model the inter-
face between the arch and the abutments and the arch and the backfill. These special ele-
ments can take into account a different frictional behaviour in the longitudinal and tangent
directions. They also allow a better description of the displacements of the contiguous ele-
ments as is the case between the arch barrel and the abutment devices ( see figure 12)
where a joint is placed.
Taking into account the compressive strength of the brick and the mortar used in the experi-
ment (154 and 1.9 MPa respectively) and using the equation in the Eurocode
f c =0.6 f b
0.65 0.25
fm (7)
fc = compressive strength of the masonry

fb = compressive strength of the bricks
fm = compressive strength of the mortar
a value of fc = 18 MPa is obtained. Using the empirical relation Ec = 1000 fc, a value of 18000
MPa is obtained for the elasticity modulus of the masonry. Note that mean and non charac-
teristic values are used for material properties as the objective of the analysis is to check the
model with the results of the test.
A non-linear analysis was done for different values of the elasticity modulus of the infill mate-
rial, ranging from 3.5 to 3.5x102 MPa. The compressive strength of the infill material has
been also changed proportionally to the Young modulus. Once the self weight and perma-
nent loads are applied, the point load at the quarter point is incremented step by step at a
rate of 12.5 kN per step. In each step the program identifies the cracked elements. A maxi-
mum of 250 iterations was fixed. Close to the ultimate load, the applied load is adapted and
limited. The program verifies that the displacements corresponding to the applied load are
tolerable, i.e. less than the imposed ones. If this is not the case, the applied load is adapted
accordingly. The absence of convergence can be interpreted as the collapse of the arch.
The summary of the results for the ultimate load and the vertical displacement of the arch
below the point of load application ( quarter point) and the horizontal displacement in the
abutment more far from the load are presented in figure 16.
35
Horiz deplacement (mm)
30
Vert deplacement (mm)
Ultimate load (*10kN) 25
20
15
10
0
3,50E+06 1,80E+07 3,50E+07 1,80E+08 3,50E+08
Infill Young modulus
Figure 16.- Ultimate load, deflection at quarter span and horizontal displacement at
abutment as function of the infill Young modulus (in Pa)
From the figure, we can see that the value of the Young modulus that better identifies the
ultimate load for the case of the limestone infill, is 3.5x107 Pa, whereas a value of 1.2x107 Pa
is the most appropriate in the case of the clay infill.
In the figures 17 to 21 are shown the principal stresses in the structure for the different val-
ues of the elasticity modulus analysed.
Figure 17.- Principal stresses for E = 3.5 MPa
Figure 18.- Principal stresses for E = 18 MPa

Finally, in figure 22 is shown the horizontal pressure on the arch for the different values of
the Young modulus of the infill material.
140,00
130,00 3.5E6,Right Side

3.5E6,Left Side
120,00
1.8E7,Right Side
110,00 1.8E7,Left Side
3.5E7,Right Side
100,00 3.5E7,Left Side
1.8E8,Right Side
90,00
1.8E8,Left Side
3.5E8,Right Side
80,00
3.5E8,Left Side
Height
70,00
60,00
50,00
40,00
30,00
20,00
10,00
0,00
-2,0E+05 -1,5E+05 -1,0E+05 -5,0E+04 0,0E+00 5,0E+04 1,0E+05 1,5E+05 2,0E+05 2,5E+05 3,0E+05
Pressure
Figure 22.- Variation of the horizontal pressure at failure on the arch barrel as a function of
the elasticity modulus of the infill material. The left side corresponds to the side where the
load is applied
From these figures and the evolution of the ultimate load and displacements with the rigidity
of the infill material ( figure 16), the following conclusions can be derived for the influence of
the arch-soil interaction:
1.- In a first stage, the vertical and horizontal displacements increase in an important way,
afterwards there is a decrease and finally there is a large increase for higher ultimate loads.
Therefore, the bridge performs quite differently depending on the stiffness of the backfill ma-
terial.
2.- For all values of the stiffness of the backfill, the failure is due to a 4 hinge mechanism.
However, as the stiffness increases, the hinges become more spread across the arch. This
can be due to the stress redistribution inside the backfill depending on the relative stiffness. 3
different behaviours can be observed:
a) For a very soft backfill, its collaboration to the resistance of the arch is negligible and the
behaviour is similar to that of an isolated arch barrel. The backfill does not transmit any load
to the soil and the arch and abutment remain fixed (no relative horizontal displacement be-
tween arch and abutment)
b) For a stiffer backfill, this gives and important resistance, enough for a relative displace-
ment between arch and abutment to appear.
c) Finally, for a very stiff backfill, the backfill does not allow the horizontal displacement of the
abutment, transmitting a high load to the soil and spreading the external load to a higher
length of the arch.
3.- The horizontal pressure in the barrel is higher in the right side, i.e., the side where the
load is not applied. According to the 3 modes of behaviour just identified (see 2), the in-
crease in the stiffness of the infill material leads to the increment of the horizontal pressure in
the arch barrel and to a spreading of the applied load to an important part of the arch. This
derives in a higher resistance of the structure.
4.- The influence of the backfill material is critical to the strength of the masonry arch. De-
pending on the stiffness one identifies 3 different behaviours: arch isolated (the contribution
of the horizontal pressure of the backfill to the resistance is negligible), arch with collaborat-
ing backfill (horizontal pressure from backfill to the arch barrel increases the arch resistance)
and arch+backfill acting as a unique structure.
5.- The horizontal displacement in the abutment increases due to the common displacement
of the arch and the abutment, up to a level stiffness where a relative displacement between
arch and abutment (sliding force is higher that the frictional force in the joint) appears and,
consequently, the displacement decreases.
Simplified model
A simple theoretical model has been calibrated with the results of the tests and the advanced
model (DIANA software). The simplified model (ARCC) is based on the classical limit analy-
sis with and static approach, and the Lower Bound and Unicity theorems. However, the origi-
nal model takes into account the compression strength (no assumption of infinite compres-
sive strength of the masonry). The numerical model is very simple and implemented in a Ex-
cel sheet. In a first step, the model without arch-infill interaction has been used to the analy-
sis of the test, showing a bad prediction of the ultimate load. Afterwards, the interaction be-
tween arch and backfill has been included in the model what has led to a better accurate
prediction of the ultimate load.
When the influence of the infill material is not considered in the arch resistance, an ultimate
load of 43 kN is obtained. This is clearly much lower than the load predicted by the advanced
model and the test result (125 kN).
The interaction of arch and infill material has been considered in the model by using the re-
sults from the advanced numerical model (DIANA). The horizontal pressure over the arch
barrel corresponding to a stiffness of 3.5x107 Pa (limestone infill) as appears in figure 23 has
been used to calculate the horizontal force transmitted from the infill to the barrel and intro-
duced in the corresponding equilibrium of forces over the arch barrel. Taking into account
this interaction, the ultimate load is 122 kN, in good agreement with the advanced model and
the test for the case of limestone infill.
140
130
120
110
3.5E7,Right Side
3.5E7,Left Side
100
Abutment
90
80
Height
70
60
50
40
30
20
10
0
-1,0E+05 -8,0E+04 -6,0E+04 -4,0E+04 -2,0E+04 0,0E+00 2,0E+04 4,0E+04 6,0E+04 8,0E+04 1,0E+05
Horizontal Pressure (Pa)
Figure 23.- Horizontal pressure (in Pa) over the arch extrados
A similar calculation has been carried out for the other values of the stiffness of the backfill
and the corresponding horizontal pressure curves shown in figure 22. The comparison of the
results obtained with the simplified ARCC model considering the barrel-backfill interaction
and the advanced model (DIANA) is presented in table 1 and figures 24 and 25.
Infill E modulus
DIANA Ultimate Load (kN) ARCC Ultimate Load (kN)
(Pa)
3,50E+08 287,5 182
1,80E+08 200 147
3,50E+07 125 122
1,80E+07 110 119
3,50E+06 55 72
Table 1.- Comparison of results from advanced (DIANA) and simplified (ARCC) theoretical
models as a function of Young modulus of infill material
350
300
Ultimate load (kN)
250
200 Diana
150 Excel
100
50
0
0,00E+00 1,00E+08 2,00E+08 3,00E+08 4,00E+08
Infill Young's modulus (Pa)
Figure 24.- Comparison of results from advanced model and simple model ARCC (Excel)
190
170
Ultimate load (kN)
150
130 Diana
Excel
110
90
70
50
0,00E+00 2,00E+07 4,00E+07 6,00E+07 8,00E+07 1,00E+08
Infill Young's modulus (Pa)
Figure 25.- Comparison of results from advanced model and simple model ARCC (Excel)
As can be seen, a good approximation is obtained with the simple model if the interaction is
considered, for values of the Young modulus of the backfill material up to 35 MPa. For higher
values, the results from the simple model are not as accurate.
The ARCC model has been used to study the influence of the variation of the most significa-
tive parameters on the ultimate load for the case of a Young modulus of the infill equal to 35
MPa. The results are shown in figures 26 to 29.
140
135
Ultimate load (kN
130
125
120
115
110
105
-35 -15 5 25
Variation of the infill density
(%)
Figure 26.- Influence of the density of the infill material on the ultimate load
210
190
Ultimate load (kN
170
150
130
110
90
70
50
-22 -12 -2 8 18
Variation of the arch barrel thickness
(%)
Figure 27.- Influence of the arch thickness on the ultimate load

129
Ultimate load (kN
127
125
123
121
119
117
115
-11 -6 -1 4 9
Variation of the span-length
(%)
Figure 28.- Influence of the span-length on the ultimate load
240
220
Ultimate load (kN
200
180
160
140
120
-16 -11 -6 -1
Variation of the rise
(%)
Figure 29.- Influence of the arch rise on the ultimate load

Modelisation of multi-ring arches

In order to investigate the capability of existing numerical models to model the behaviour of
multi-ring arches when loaded statically up to failure, the results from those models are
checked against the results from a model test.
Description of the test
The models corresponds to the bridge 3-2 tested by Gilbert and Melbourne [18]. The bridge
geometry is presented in figure 30.
Figure 30.- Main dimensions of the bridge 3-2 tested by Gilbert and Melbourne [18]
The span to raise ratio is 4:1. The load was applied at the quarter of span. The two rings are
separated by damp sand rather than mortar to simulate the defect of ring separation and the
spandrel walls were detached in order to simulate also this defect in the bridge. The materi-
als used were Class A Engineering bricks in conjunction with a weak 1:2:9 (ce-
ment:lime:sand) mortar. The backfilling was made by graded crushed limestone compacted
in layers. The mean densities of the brickwork and backfill were measured as 2310 Kg/m 3
and 2260 Kg/m 3 . The bridge was loaded at quarter span. The failure load was 360 kN.
Description of the numerical model

The FEM developed with DIANA is presented in figure 31 and comprises 7000 triangular
elements and 200 special elements to model the interface between the rings. As the bricks
and mortar used in this test were similar to the test presented above, a compressive strength
of 18 MPa and Young modulus 18000 MPa were taken for the brickwork. Because crushed
limestone was also used in the present test, a value of 35 MPa was also considered for the
Young modulus of the backfill (similar to the previous test of single arch with limestone back-
fill).
Figure 31.- 2-D FEM model
Results of the numerical model

In figure 32 is shown the stress state in the arch for different levels of the applied load. One
may distinguish the different behaviour of the two rings and the separation between them for
a certain load level. At the ultimate load, the distribution of reactions in the abutment for both
rings is presented in table 2.
Figure 32.- Principal stresses for increasing level of the applied load
Vertical Reaction Horizontal Reaction

Abutment closest to loading 72.2 % 68 %
Other abutment 27.8 % 32 %
Upper ring 60.9 % 71.9 %
Lower ring 39.1 % 18.2 %
Table 2.- Distribution of reaction between abutments and rings
As shown in the example, DIANA accurately predicts the ultimate load where the ring sepa-
ration takes place and the stress levels due to service loads. Also the model is able to predict
the load level where the ring separation will take place according to the load increase. How-
ever, from the works developed so far, it seems that simplified models as the Ring software
do not predict as accurate estimates for the ultimate load. In this case, the ultimate load re-
sulting from the RING software is 320 kN, something lower than the real 360 kN (figure 33).
Moreover, in the RING method, the assumption of ring separation has to be introduced in the
analysis from the beginning and is not obtained from the loading application process.
Therefore, at this stage and for multi-ring arches, it will be better to concentrate on the possi-
bility of using simplified models only to predict the stresses in the arch due to service loads.
First of all, the possibility of a linear elastic model should be explored, comparing their results
with those coming from more advanced models as DIANA.
Figure 33.- Failure of the arch predicted by RING
Influence of arch skewness
Several tests have shown that a very influencing parameter in the load capacity of masonry
arch bridges is the skewnes at the supports. The main objective of the study herein pre-
sented is to investigate such influence and also to see the capability of existing theoretical
models to simulate the real behaviour of skewed arch bridges. Two theoretical models are
investigated. The first is based on a 3D solid FEM using software DIANA and the other is
based on a generalized advanced beam element using the software CRIPTA [4]
To validate the capacity of the FEM model in the analyses of skewed masonry arches, the
models tested by Melbourne and Hodgson [22] with angles of 0, 22.5 and 45 without span-
drel walls were selected (Table 3). These bridges spanned 3 m. The arches were segmental
with a maximum rise at midspan of 0.75 m and its depth was constant of 0.22 m. All arches
present the same supported length of 2.84 m (Figures 34 and 35).
Skew angle Span Skewed span

0 3m 3m
22.5 3m 3.25 m
45 3m 4.24 m
Table 3.- Skew, span and skewed span of the bridges used to validate the model
Figure 34.- Geometry of bridges with 0 and 22.5 skew angles (m).
Figure 35.- Geometry of 45.0 bridge (m).

Loads acting on the bridge were its dead weight, including the spandrels infill and an exter-
nal line load (knife) at L/4 applied parallel to the abutments. The latter load was increased up
to reaching the failure. The spreading of the external loads through the infill was simulated by
a triangular distribution with an angle of 1H:2V. Boundary conditions at the abutments were
supposed perfectly rigid, therefore, all movements (degrees of freedom) were blocked.
A mesh of 3D solid (brick) elements were used for the arch: a ten nodes isoparametric tet-
raedric element was selected (Figure 36).
Figure 36.-. Tetraedric isoparemetric FE (CTE30 in Diana code)
Models contained 12000 elements with a resulting amount of 20000 nodes (Figures 37 and
38). Dimensions of the mesh were determined by the depth of the arch and the length and
wideness of the arch. A critical aspect of the mesh was selecting the number of finite ele-
ments across the depth. That number has to be enough to allow huge gradients of deforma-
tion where hinges had to appear and rotate. However, total number of nodes (and degrees
of freedom) had to be controlled to have calculations within a reasonable time. After some
parametric studies, the depth of the arch was modelled by five elements and, thus, eleven
nodes. All models have eleven nodes in its depth and have varying number of nodes in longi-
tudinal and transverse direction (see table 4).
Figure 37.-. Finite element mesh of models of 0 and 22,5 skew.
Number of elements Number of nodes

Skew
Long x transv x depth Long x tranv x depth
0 30 x 20 x 5 61 x 41 x 11
22.5 37 x 25 x 5 75 x 51 x 11
45 28 x 44 x 5 57 x 89 x 11
Table 4.- Number of elements and nodes of the models.
Figure 38.- Finite element mesh of model of 45 skew.
MATERIAL MODELLING
Masonry has been suposed to present isotropic mechanical properties and smeared crack-
ing. Past experiencies demonstrate that such approach is enough accurate when looking for
global plastic mechanisms in arches.
A total strain crack model has been selected, based on the Modified compression field theory
proposed by Vecchio and Collins, including the extension to 3D develped by Selby and Vec-
chio. The Total Strain crack model evaluates stresses in the directions defined by cracks.
Exponential softening was selected for cracking in tension. Figure 39 shows the stress strain
relationship after cracking in tension.
Figure 39.- Softening in tension.
Which expression is:
cr
c
nncr ( nncr ) nn 0 < nn cr

< nn
cr
= 1 cr
si
.ult
ft nn.ult si nn.ult < nn <

cr cr

0
where c = 0.31
Factor is:
x = x =1 x = x =1 c
= y ( x)dx = y ( x)dx + 0dx = (1 x c )dx =
x =0 x =0 x =1 x =0 1+ c
Then = 0.23664122 and the ultimate deformation in tension is:
G If
cr
nn .ult = 4.226
hf t
Constant shear retention was used (Figure 40)
Figure 40.- Shear retention.
In compression, a perfect elastoplastic behaviour was choosen. In spite of being quite simple
it is a very good approach and improves the speed of calculations (Figure 41).
Figure 41.- Stress strain diagram in compression.
Table 5 shows the material properties used in the analysis. High value of ultimate tension
strain was required to allow the formation of a full failure mechanisms.
Material properties
Deformation modulus E (kN/m2) 9000 Experimental
Poisson modulus 0.2 Experimental
Tensile strength ft (N/mm2) 0.05 Experimental
Compressive strength fc (N/mm2) 28.0 Experimental
Fracture energy Gf (Nm/m2) 100
Ultimate tension strain u 0.01
Table 5.- Material properties used in the analysis.
Full Newton-Raphson algorithm was used to solve the nonlinear problem. Convergence tol-
erances were 0.01 in relative value for both nodal displacements and nodal forces.
RESULTS
Results are presented for each skew.
0 skew
Figure 42.- Vertical displacements of the arch subjecte to dead load (m).
Figure 43.- Deformed shape of the arch subjecte to dead load (x6000).
Figure 44.- Normal stresses under DL. in N/m2. Compressive stresses are negative.
The ultimate load achieved was 213 KN with a collapsing mechanism of four hinges (figures
45 and 46). Compression stresses are less than the compressive strength of the masonry
(28 N/mm2).
Figure 45.- Stress state at failure (N/m2).

Figure 46.- Deformed shape at failure. (x130).
Figure 47 shows the evolution of the vertical displacement at quarter-span under the increas-
ing live load.
SKEW 0
250
200
Applied load (KN)
150
100
50
0
0 0,2 0,4 0,6 0,8 1 1,2
Displacement (mm)
Figure 47.- Load vs. displacement under the live load diagram.
22.5 skew
Under dead loads, compressive stresses are concentrated at the obtuse abutments. The
peak value is -0.55 N/mm2 (Figure 48). Figure 49 shows the distribution of horizontal force at
the springings.
Figure 48.- Stress state N/m2 (dead loads)
Figure 49.- Forces in X axis at the springings of the arch subjected to DL.
Maximum vertical displacement under DL is 0.06 mm (Figure 50) at the free sides of the
structure.
Figure 50.- Vertical displacements under DL (in m).

Figure 51 presents the compressive stresses and cracking at different levels of live load in
both intrados and extrados.
Figure 51.- Stress state of the intrados and the extrados of the arch in N/m2
at 40%,60% and 80% of the ultimate live load.
Ultimate live load was 202 KN (5% less than the 0 skew arch). Figure 52 shows the evolu-
tion of the vertical displacement under the increasing live load.
SKEW 22.5
250
Axle load (KN) 200
150
100
50
0
0 0,2 0,4 0,6 0,8 1 1,2
Vertical displacement (mm)
Figure 52.- Load vs. displacement under the live load diagram.
The overall plan distribution of vertical displacements obtained by FE model of the bridge
with 22.5 skew presents good agreement with those registered in the test (Figure 53). Its
values are not directly comparable because only the dead weight of the infill was accounted
for in the FE model and not the horizontal pressure
Figure 53.- Vertical displacement at failure in m.

45 skew
Under DL, this bridge presents larger vertical displacements and compressive stresses (Fig-
ure 54). Table 6 shows a comparison of such results obtained in the three models analyzed.
Figure 54.- Vertical displacement under DL in m.
Stresses at the springings Stresses at the springings

Vertical dis-
Extrados Intrados
skew placement
(mm) max (N/mm2) min (N/mm2) max (N/mm2) min (N/mm2)
0 0.0520 -0.100 -0.150 -0.250 -0.500
22.5 0.0690 -0.005 -0.450 -0.250 -0.550
45 0.1600 0.010 -0.550 -0.400 -1.000
Table 6.- Vertical displacement and stresses of the bridges under Dead Load obtained by FE
models.
Figure 55.- Stress state of the intrados of the arch in N/m2 at different values of the ultimate
live load.
Figure 55 shows the evolution of the stress state of the intrados of the bridge subjected to an
axle live load. It is worth noting that the formation of the hinge at the springing more distant
from the axle load is more complex than in the other bridges. Figure 56 present the cracks
and damages at the extrados reported by Melbourne with the stress state at failure obtained
in this work. There is a good agreement between maximum compressive stresses and
cracked zones of the numerical model with those from the test.
Figure 56.- Extrados at faiure for 45 skew.

Figures 57 and 58 illustrate the inhomogeneity of stresses and cracking at different cross
sections parallel to the springings and parallel to the arch sides, respectivelly.
Figure 57.- Stress state at failure in different cross sections.
Figure 58.- Stress state at failure in different cross sections parallel to the arch side.
The overall plan distribution of vertical displacements obtained by FE model of the bridge
with 45 skew presents good agreement with those registered in the test (Figure 59). The
values are not comparable because only the dead weight of the infill was accounted for in the
FE model. Ultimate load obtained by the model was 104 KN and the experimental 337 KN.
Figure 59.- Vertical displacement at failure in m. Comparison with test results

VARIATION OF THE SKEW
Some parametric studies had been developed using 3D solid finite element models and so-
phisticated beam elements (CRIPTA) [4]. Results were also compared with those provided
by the experiments developed by Melbourne.
A total of eleven different skew angles were used to analyse the effect on the ultimate load
using both modelling techniques. Skew varies from 0 to 56.25. In fact, a skew of 60 was
numerically tested but it was unable to resists its dead load. Table 7 and Figure 60 summa-
rize the results.
ULTIMATE LOAD
Distance Experimental 3D solid Sophisticated

Skewed FE model
Skew between test (Mel- beam model
span
abutments bourne) (DIANA) (CRIPTA)
0 3m 3m 540 KN 213 KN 205 KN

5.625 3m 3.01 m - - 205 KN
11.25 3m 3.06 m - 210 KN 205 KN
16.875 3m 3.13 m - - 200 KN
22.5 3m 3.25 m 540 KN 202 KN 200 KN
28.125 3m 3.40 m - - 190 KN
33.75 3m 3.60 m - 150 KN 190 KN
39.375 3m 3.88 m - - 185 KN
45 3m 4.24 m 337 KN 104 KN 185 KN
50.625 3m 4.73 m - - 185 KN
56.25 3m 5.40 m - 40 KN 185 KN
Table 7.- Failure loads achieved by different models varying the skew.
Effect of the skew
CRIPTA DIANA MELBOURNE
600
500
400
Axle load (KN)
300
200
100
0
0 10 20 30 40 50 60
Skew angle (in )
Figure 60.- Failure loads varying the skew.
Experimental results [22] and numerical results of Table 7 and Figure 60 show that the load
capacity of skewed bridges with an angle from 0 to 22.5 is almost the same of the right one.
The sophisticated beam model was unable to yield accurate results when increasing the
skew. Table 7 and Figure 60 show that ultimate loads are systematically underestimated
when compared with those of the experimental tests. That is because the constraints intro-
duced by the infill to the movements of the arch are neglected, despite taking into account
the weight and the spreading of the live load were accounted for the infill. However, tenden-
cies shown by experimental tests are very well reflected in the analysis of skewed arches
without simulating the stiffness and strength of the infill. In fact, there is proportionality be-
tween experimental and numerical results varying the skew (Table 8)
Skew 0 45 load capacity

Melbourne 540 KN 337 KN -40%
DIANA 213 KN 104 KN -50%
Table 8.- Comparison of the ultimate loads.
VARIATION OF THE AXLE LOAD POSITION

In this parametric study, the worst position of the axle load is analysed in two different
skewed arch bridges using different techniques. Bridges with a skew of 22.5 and 45 are
analysed to find out the worst loading position.
Both 3D solid FE (Diana code) and sophisticated beam model (CRIPTA) are used in this
parametric analysis. All results presented in Figures 61 and 62, corresponding to 22.5 and
45 skew agree that the worst loading position was L/4. Figure 63 presents in an amplified
scale the results provided by 3D solid FE in the bridge with 45.
SKEW 22.5
CRIPTA DIANA
1800
1600
1400
Failure load (KN)
1200
1000
800
600
400
200
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
Figure 61.- Diagram Load vs. Position on the arch with a 22.5 skew.
SKEW 45
CRIPTA DIANA
1600
1400
1200
Failure load (KN)
1000
800
600
400
200
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
Figure 62.- Diagram Load vs. Position on the arch with a 45 skew.
SKEW 45
DIANA
300
250
Failure load (KN)
200
150
100
50
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
Figure 63.- Diagram Load vs. Position on the arch with a 45 skew.
As a reference, detailed analysis of the worst position for the axle load on the right bridge is
presented in Figure 64. The results were obtained by using the sophisticated beam model
(CRIPTA).
SKEW 0
CRIPTA
2000
1800
1600
Failure load (KN)
1400
1200
1000
800
600
400
200
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
Figure 64.-Diagram Load vs. Position on the right bridge.

All results show the high load capacity of loading these arches at mid-span. Such result was
well known in the case of right bridges, where a five hinge mechanism is required when sub-
jected to this symmetric loading. The analysis performed show that skewed bridges have the
same behaviour when subjected to axle load applied at mid-span.
CONCLUSIONS
Experimental results [22] and numerical results herein presented show that the load
capacity of skewed bridges with an angle from 0 to 22.5 when failing by the 4-hinge
mechanism is almost the same of the right one. However, compression stresses in
the springings are concentrated at the obtuse angle area in the case of skewed
bridges.
Failure mechanisms of arches presenting and skew over 22.5 are fully 3D.
From the examples analysed can be concluded that, in the case of maintaining con-
stant the distance between supports, varying skew angle do not change the worst po-
sition of the axle load on the platform. So, the worst position of the load can be de-
duced in a simple model.
As it is well known, the effect of the infill constraining the movements of the arch af-
fect significantly the load capacity. In the numerical examples presented, only the
weight and the spreading of the live load were accounted for the infill. So that, load
capacities were systematically underestimated. However, tendencies shown by ex-
perimental tests are very well reflected in the analysis of skewed arches without simu-
lating the stiffness and strength of the infill. In fact, there is a clear proportionality be-
tween experimental and numerical results varying the skew.
With the models used, 3D nonlinear finite element calculations for skew arches were
highly time-consuming.
As could be foreseen, sophisticated beam elements were not capable to fully simu-
late the ultimate capacity of masonry arch bridges with large skew, despite simulating
complex formation of hinges. Results on bridges with a skew between 0 and 22.5
present very good agreement with those provided by 3D solid FE.
5 Basic variables
For a probabilistic approach to the assessment of masonry arches, the statistical definition of
the basic variables is mandatory. The statistical definition of the basic variables should be
based on the available data existing in the literature up-dated with the measurements ob-
tained directly on the bridge under assessment. The up-dating can be carried out using
Bayesian techniques with the data available from the inspection and monitoring in the bridge.
The most important data that needs to be collected in order to improve the quality of the pre-
dicted carrying capacity using probabilistic techniques is shown in the following.
5.1 Dimensions
The most representative dimensions for a probabilistic analysis of masonry arches are the
arch thickness, arch width, depth of fill at crown, rise at mid-span and span-length. At
this time, there is not enough data available to obtain the statistical definition of these vari-
ables and in the literature information how to get the statistical parameters is not available
too. Based on some measurements made by the authors in several masonry arch bridges, a
rough estimate of 10 % coefficient of variation for the arch thickness and 5 % for the arch
width can be assumed. The distribution function can be assumed as Normal. Because these
values have been obtained from a very limited number of samples, it is recommended that
up-dating of this information should be done from results gathered in the specific bridge un-
der assessment.
5.2 Materials
The most representative material properties for a probabilistic assessment of a masonry arch
are the compressive strength of masonry, tensile strength of masonry, compressive
strength and modulus of elasticity of filling material, , density of masonry and filling
material, endurance limit to fatigue of masonry and shear strength (interface masonry-
mortar). At the present time there is not available data in the literature on the statistical dis-
tribution and parameters of these properties.
5.3 S-N diagrams. Probabilistic fatigue models

Normally, the characterization of fatigue properties of materials is done through the S-N dia-
grams. As previously explained, fatigue failure may occur either under the combined com-
pression-bending action (as in the case of single ring arches) or because of the shear action
in the interface unit-mortar (case of multi-ring arches). Because the two mechanisms can be
of different origin, a separation is made between them and the corresponding S-N curves for
fatigue in each case is derived.
5.3.1 S-N diagrams for masonry in compression-bending

As presented in 4.3.1, assuming a Weibull distribution, a complete definition of the fatigue
strength in probabilistic terms requires the knowledge of the two parameters of the distribu-
tion for all possible stress levels.
A simple way for obtaining the parameters of the Cumulative Distribution Function (CDF) for
all stress levels was presented in [14] for the fatigue analysis of concrete based on the as-
sumption of a constant standard deviation of the number of cycles to failure for all stress lev-
els. The same method was applied by Crespo-Minguillon and Casas in [11] in the case of
fatigue strength of prestressing and reinforcing steels in structural concrete. The expression
of the S-N curves for concrete and steel are very similar, in the form:
m

Concrete: N c max =K (8)
f cm
Steel: N m = K (9)
where cmax is the maximum applied stress, fcm is the compressive strength of the concrete,
N is the number of cycles to failure, is the applied stress increment, m and K are the char-
acteristic power and constant associated to a specific branch of the S-N curve. In equation
(8) it is assumed that the minimum stress level is zero and therefore the stress range is equal
to cmax
Assuming that fatigue phenomena causing crushing of mortar and brick subjected to com-
pression is similar to that of concrete, Hwans methodology could be adopted here to define
also the fatigue strength of masonry. Assuming a Weibull distribution, and noting that s is a
constant deviation of log N for a given log , for all stress increments, the parameters of the
distribution could be estimated through the following expressions [11]:
2
2 = (10)
6s 2
0.5772
ln u = + ln( K m ) (11)

The values K and m can be obtained from the available fatigue tests on masonry samples in
compression. Normally, a linear regression through experimental data reported in a log N-
log graph is used.
However, the direct application of Hwans proposal to masonry structures is highly debatable
mainly for two reasons:
1.- The fatigue behaviour of the material may be sensible not only to the stress range applied
to the material, but also to the minimum level of stress imposed during the cycle loading.
2.- It has to be demonstrated in some way that the assumption of a constant value of s (what
according to equation (10) implies a constant value for the parameter ) independent from
the stress level is valid.
The only experimental data in masonry specimens tested to fatigue in different conditions is
available in [9]. They conclude that the high cycle fatigue strength of wet and submerged
brick masonry test specimens is only slightly less than the fatigue strength of similar labora-
tory dry test specimens and propose the following equation for the lower bound fatigue
strength:
(SS max ) 0.5

F (S ) = =0.7 0.05 log N (12)
Su
where F(S) is the so-called function of induced stresses. S and Smax are the induced stress
range and maximum induced stress respectively and Su is the quasi-static compressive
strength under similar loading conditions and N is the number of constant amplitude load
cycles. As shown in equation (12), the fatigue strength depends not only on the stress range
but also on the maximum and minimum level of stress in the material.
In order to obtain the parameters of the statistical distribution of the number of cycles to fail-
ure, one has to perform a statistical analysis of fatigue data at every stress level S. It is very
difficult to do this directly when the stress level and the stress range are both variables. For
this reason, the variable F(S) as defined in equation (12) was first adopted in [9] as indicative
of the stress level in the material.
According to the expression of the Weibull distribution with two parameters ( equation (4)
with n0 = 0)
n
FN (n) =1 exp , n 0 (13)
u
and considering the survival function, SF, as
SF(n) = P (N > n) = 1- FN (n) (14)
taking the logarithm in both sides, it becomes:
ln [ln (1/SF)] = ln(n)- lnu (15)
This equation can be used to verify if the experimental data obtained from a series of tests
follows the Weibull disitribution. From the experimental data, the survival probabilities SF for
each fatigue life N can be calculated according to SF = 1-[m/(k+1)] where k = total number of
N data for a certain value of the stress level FS, and m = the ascending order of the N data
point considered. By means of a linear regression to the experimental values, the parameters
and u can be calculated.
From the data available in [9], the figures 65 to 69 were obtained. They draw the regression
analysis for different values of the function of induced stresses, F(S).
FS=0.70
1
y = 1,0565x - 9,668
R2 = 0,6873
0,5
0
0 2 4 6 8 10 12
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)
Figure 65.- Regression analysis for proposed Weibull distribution. FS = 0.7
FS=0.65
1
y = 0,2666x - 3,2458
2
R = 0,8255
0,5
0
0 2 4 6 8 10 12 14 16
-0,5
ln(ln(1/SF)
Serie1
Lineal (Serie1)
-1
-1,5
-2
-2,5
ln(N)

FS=0.6
1
y = 0,3524x - 4,4408
R2 = 0,8089
0,5
0
0 2 4 6 8 10 12 14 16
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)
FS=0,55
0,5
y = 0,4386x - 6,3641
2
R = 0,8936
0
0 2 4 6 8 10 12 14 16
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)

FS=0.50
0,5 y = 0,5006x - 6,9347

2
R = 0,9547
0
0 2 4 6 8 10 12 14 16
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)
As can be observed from the figures, the obtained correlations are not very good. For this
reason, according to the results reported in [15], where they propose a probabilistic fatigue
model for plain concrete, the following fatigue formula is proposed to take into account the
minimum value of repeated stress:
max
log S = log =1 b(1 R ) log N (16)
u
S is the ratio of the maximum loading stress to the strength and R is the ratio of the minimum
stress to the maximum stress min / max. In order to work with only one stress variable in-
stead of two, the introduction of the equivalent fatigue life EN is done. EN is defined as
EN=N(1-R). In this way, equation (16) takes the form of the Wholer equation (logS= a-b log
EN).
In the following figures are presented the results of the linear regression of the experimental
data from [9] with the new variables S and EN instead of FS and N
S= 0.9
0,5
y = 0,8511x - 5,336
R2 = 0,9603
0
0 1 2 3 4 5 6 7 8
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(EN)
Figure 70.- Regression analysis for proposed Weibull distribution. S = 0.90
S=0.8
y = 0,5353x - 4,3585
R2 = 0,9772
0,5
0
0 1 2 3 4 5 6 7 8 9 10
ln(ln(1(SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(EN)

S= 0.75
y = 1,0753x - 7,7298
2
0,5 R = 0,9746
0
0 1 2 3 4 5 6 7 8 9
-0,5
ln(ln(1/SF))
Serie1
Lineal (Serie1)
-1
-1,5
-2
-2,5
ln(EN)
S=0.70
y = 0,4604x - 4,8822
2
R = 0,8753
0,5
0
0 2 4 6 8 10 12 14
-0,5
ln(ln(1/SF))
Serie1
Lineal (Serie1)
-1
-1,5
-2
-2,5
ln(EN)

S=0,65
0,6
0,4
y = 0,2379x - 2,521
0,2
R2 = 0,5046
0
0 2 4 6 8 10 12 14
-0,2
ln(ln(1/SF))
Serie1
-0,4
Lineal (Serie1)
-0,6
-0,8
-1
-1,2
-1,4
ln(EN)
S=0.60
0,6
y = 0,4202x - 5,3679
0,4 2
R = 0,9703
0,2
0
0 2 4 6 8 10 12 14 16
-0,2
ln(ln(1/SF))
Serie1
-0,4
Lineal (Serie1)
-0,6
-0,8
-1
-1,2
-1,4
ln(EN)

S=0.55
0,6
0,4
0,2
y = 0,8785x - 9,0308
R2 = 0,8778
0
8,6 8,8 9 9,2 9,4 9,6 9,8 10 10,2 10,4 10,6
-0,2
ln(ln(1/SF))
Serie1
-0,4
Lineal (Serie1)
-0,6
-0,8
-1
-1,2
-1,4
ln(EN)
As seen in the previous figures, the correlation is much better than in the previous case.
Therefore, equation 16 is proposed in the present study as the most appropriate S-N curve
for masonry subjected to compression. From the regression analysis, the corresponding pa-
rameters of the Weibull distribution for different values of S are presented in table 9.
S lnu u r2 r
0.9 0.8511 5.336 528 0.96 0.98
0.8 0.5353 4.3585 3436 0.98 0.99
0.75 1.0753 7.7298 1324 0.97 0.98
0.70 0.4604 4.8822 40306 0.88 0.94
0.65 0.2379 2.521 40010 0.50 0.71
0.60 0.4202 5.3679 353144 0.97 0.98
0.55 0.8785 9.0308 29138 0.88 0.94
Table 9.- Parameters and u of Weibull distribution for different values of stress level S
As can be seen from table 9, the regression coefficients for all stress levels are very high
except for the case S = 0.65. Also, from table 9, it becomes clear that the hypothesis of con-
stant value of deviation s ( or constant value of ) for different stress levels is not valid.
The parameters and u in table 9, depending on the stress level, should be used for the
reliability-based assessment to fatigue of existing masonry arches. In fact, as Roberts et al.
point out in their paper [9], the low compressive strength of the brick and mortar used to
manufacture the test specimens were intended to be representative of the type of bricks and
mortar likely to be encountered in relatively old masonry arch bridges.
Depending on the number of cycles of load N for each stress level S present in the loading
histogram of the bridge, either the reliability index to fatigue failure and/or the remaining ser-
vice life with a certain confidence level can be obtained.
Based on these results, and the fatigue equation of the type:
S = A x N-B(1-R) (17)
the values in table 10 are obtained for the coefficients A and B for different values of the sur-
vival function SF
SF A B r
0.95 1.106 0.0998 0.95
0.90 1.303 0.1109 0.98
0.80 1.458 0.1095 0.97
0.70 1.494 0.1023 0.93
0.60 1.487 0.0945 0.90
0.50 1.464 0.0874 0.86
Table 10.- Parameters of fatigue equation depending on the required confidence level
The mean value of the coefficient B for SF>0.6 is 0.1034.

In figure 77 are presented the experimental data (divided in dry, wet and submerged speci-
mens and the run outs are also indicated) and the linear functions for a survival probability
SF = 0.95 and 0.7 according to the coefficients A in table 10 and the mean value of B =
0.1034. In summary, the proposed fatigue equation (S-N relation) for a survival probability of
95 % for masonry in any condition (dry, wet or submerged) is:
S = 1.106 N-0.1034(1-R) S >0.5 (18)
An endurance limit can be observed for S = 0.5

The advantage of equation (18) compared to the one proposed in [9] (see equation 12) is
that it fits better the experimental data and has behind a reliability-based background that
allows to define a survival probability or confidence level. Additionally, the equation is similar
to the one used for plain concrete only with different values for coefficients A and B. There-
fore, a similar equation with different parameters can be used both for concrete and masonry
under compression.
All data
0
0 1 2 3 4 5 6 7
-0,1
-0,2
Dry
wet
-0,3 submerged
SF=0.95
log S
SF=0.7
-0,4 Run out
Lineal (SF=0.95)
Lineal (SF=0.7)
-0,5
-0,6
-0,7
log (EN)
Figure 77.- Experimental data (all tests) and proposed fatigue equation for probability of fail-
ure of 5 and 30 %.
5.3.2 S-N diagrams for masonry in shear

No experimental data from laboratory specimens is available at the moment to derive the
corresponding S-N relationships as shown in 5.3.1 for the case of masonry in compression.
The only available data correspond to cyclic loading tests carried out at the University of Sal-
ford [1] in multi-ring arches failing by ring separation. This case may be considered as repre-
sentative of the brick joints subjected to shear action.
SN curve
Anticipated ISN curves
100
Mec hanism Ring separation
Mechanism
Load (% of static load)
Ri Ring
ng s epar
se at ion
pa
rat
i on
10
5m 'strong' arch
3m 'strong'arch
5m 'weak' arch
1
1
10
100
1,000
10,000
100,000
1,000,000
10,000,000
Number of cycles
Figure 78.- Results of laboratory tests for high cycle loading [1]
Figure 78 is a summary of the results obtained in the Salfords tests. As shown there, the
fatigue mechanism of ring-separation can also be represented by an straight line in a log S
versus log N diagram. The diagram in figure 78 was obtained from tests carried out in the
whole arch and not in small specimens. For this reason, despite the materials for the 5 and 3
m strong arches are the same, a different behaviour in fatigue is observed depending on the
archs span. The main difference between both arches is the span-length and, as a conse-
quence, the stress level in the masonry for the case of self-weight plus permanent load
(minimum stress) and the stress level due to the applied test load (maximum stress). As dif-
ferent stress levels during the tests lead to different S-N diagrams, we may assume, as in the
case of the masonry subjected to compression, a stress versus number of cycles to failure
relationship in the form:
S = A N-B(1-R) S >0.4 (19)
The two different straight lines in figure 78 for the strong brick and the 3 m and 5 m span can
be, then, explained by a different value of R (stress minimum/stress maximum) in the tests in
the 3 and 5 m span arches.
The results from the tests in Salford show and endurance limit around 40 % of the maximum
static load for 3 m span arches and good quality of brick. A similar endurance limit can be
defined in the case of 5 m span arches made of weak brick.
Therefore, based on the very few data available, it seems that a similar format of the S-N
diagram can be assumed for the masonry in compression and in shear. Only the coefficients
A and B will be different. As presented in the previous chapter, the values of A and B have
been derived in this work for fatigue of masonry in compression. To obtain these coefficients
in the case of shear, we propose a set of tests similar to those reported in [9], but where the
predominant action is the shear stress in the mortar between bricks. The idea is to test small
specimens and not complete arches as the last is much more difficult and expensive. Tests
on a representative number of specimens will allow to define if a Weibull distribution also fits
well the experimental results in this case, and to obtain the parameters of the distribution.
From this, the procedure of fatigue analysis as described in chapter 4.3, based on Miners
rule, can be applied.
Up to the moment when the results of the tests of masonry in shear may be available, the
proposal is to use some rough estimates of parameters A and B based on the fatigue tests in
full arches carried out at Salford University and described in [1]. According to equation 19
and the experimental values shown in figure 47, we can obtain the following relationships:
a) 3 m strong arch
N = 1x106 log A B (1-R1)x 6 = log 0.25

N = 23500 log A B (1-R1)x 4.37 = log 0.48
b) 5 m strong arch
N = 1x106 log A B (1-R2)x 6 = log 0.62

N = 45700 log A B (1-R2)x 4.66 = log 0.72
According to the data from the tests, we may roughly assume the values R1 = 0.42 and R2 =
0.56, what gives the following results: A= 2.69, B = 0.3 for the 3 m arch, and A = 1.20, B=
0.11 for the 5 m arch.
Taking into account the few experimental data available and until more experimental data is
obtained, the fatigue equation proposed is:
S = 2.69 N-0.3(1-R) S >0.4 (20)
The methodology to apply for the assessment of multi-ring arches to fatigue will be similar to
that proposed for single-ring arches. In this last case, it is necessary to have models to de-
rive the compression in the most critical points of the arch as a function of time depending on
the external variable actions. In the same way, in the former case, the availability of models
to obtain the shear stresses in the contact between rings will be necessary to perform a ser-
vice life analysis. However, because the actions due to the passage of the normal trains will
not cause a high stress level, even a linear-elastic model can be used to calculate the com-
pression and shear stresses at defined critical points of the arch.
Depending on the span-length of the arch, the number of cycles of load to take into account
will be the total number of axles or the total number of trains passing the bridge. The vibra-
tion amplitudes due to dynamic effects of the train loads will be very small for this type of
bridges due to the damping through the infill, and, therefore, can be neglected in the calcula-
tion as they will be under the endurance limit.
6 Examples
6.1 Assessment of ULS by 4-hinges mechanism
The example of application corresponds to the Magarola river bridge. The Magarola bridge is
a masonry arch bridge located near Barcelona in one of the busiest highways in Spain, con-
necting the city with Spains capital, Madrid. The structure is a multi-span arch bridge with 5
arches of 20 m each. A probabilistic capacity assessment is carried out, starting with the ex-
perimental testing program in order to get information about the geometric dimensions and
the strength characteristics of the materials, as no documentation is available at all. This is
normally the case of most existing masonry arches. The safety of the bridge is obtained in
terms of the probability of failure (reliability index). The safety level is evaluated for the actual
existing traffic which is available from the traffic surveys and using theoretical models with
different level of complexity and accuracy (from linear elastic at section level to non-linear at
system level).
Despite being a highway bridge, the methodology presented here is of full application to rail-
way bridges, only changing the characteristics of the live load.
6.1.1 Description of the bridge

The Magarola bridge is located in Esparreguera, a city close to Barcelona, on the N-II high-
way and was built in the IXX Century. The condition assessment carried out before the ca-
pacity assessment concluded that the bridge is in good condition despite the enlargement of
its roadway with a concrete slab made in 1990 that carries two lanes of traffic in one direc-
tion. The bridge (Figures 79 and 80) is composed of five brick masonry barrel vaults span-
ning twenty meters. The overall length of the structure is 160 m and its maximum height is 25
m. The section of the piers is variable and measures 7x3 m at the abutments of the arches.
Figure 79.- Lateral view of the Magarola bridge
The piers are built of rubble stone masonry reinforced with ashlars in their corners.
Figure 80.- Plan and front view of Magarola bridge.

Because no drawings or other documentation was available, all the data used in the study
were obtained during the geometric and mechanical characterisation of the bridge carried out
in April 1997. The main geometric parameters of the arches and the mechanical properties of
the materials derived from the testing program are shown in Tables 11and 12 respectively.
The shape of the arches was defined by obtaining the co-ordinates of twenty-one (21) points.
This made possible to prove the exactitude of the semi-circumference they describe. Six ver-
tical boreholes of 2.10 m to 2.25 m (Figure 81) and four horizontal ones of 1.5 to 3.17 m
(Figure 81) allowed for recognition of the inside geometry of the bridge and its composition,
and obtaining several cores to be tested in the laboratory (Figure 82). Some of the tested
cores, which included brick and mortar joints, gave valuable information about the masonrys
strength and the characteristics of the filling material. Horizontal boreholes detected concrete
backing on the four piers.
Figure 81.- Extraction of vertical and horizontal cores
Figure 82.- Column of the materials extracted from a vertical hole ( pave-
ment+concrete+backfill material+ masonry)
Arch shape Semicircular (barrel)

Free span 20.00 m
Arch thickness 1.11 m
Arch width 7.00 m
Rise at mid-span 10.00 m
Depth of fill at crown 1.15 m
Arch ring material Brick masonry
Infill material Gravelly sand/brown
clay
Backing Plain concrete
Table 11.- Results from the geometric survey
Arch ring
Compressive (MPa) (experi- 15.0
Strength ment)
Deformation (MPa) (assumed) 3000
modulus
Tensile strength (MPa) (assumed) 0.01
Density (kg/m3) (experi- 1800
ment)
Fill
Compressive (MPa) (assumed) 1.0
Strength
Deformation (MPa) (assumed) 30
modulus
Density (kg/m3) (experi- 2300
ment)
Table 12.- Average material properties derived from the tests or assumed
6.1.2 Assessment assuming linear-elastic behaviour

The failure mechanism adopted in a first step is the one defined in 4.1.2. Failure occurs when
in the most critical cross-section of the arch, the maximum compression stress exceeds the
maximum allowable stress in the masonry material, or when the compression block (x) in the
masonry is not able to balance the axial load (N) and bending moment (M) due to the exter-
nal actions. N and M are calculated assuming a linear-elastic behaviour of the arch. The
Limit State Function is equation (1). From the geometrical survey performed in different parts
of the bridge and the cores extracted from the masonry, the values of the variables and their
variability as appear in table 13 were deduced. Also in the table are displayed the values
obtained for N and M at the most critical sections using a linear-elastic model
Variable Mean C.O.V. (%) Type

H (m) 1.11 10 Normal
B (m) 7.0 5 Normal
Fc (MPa) 15.0 20 Normal
Nsw+pl (kN) 1 875 5 Normal
2 3757
3 3129
NLL (kN) 1 412 10 Gumbel
2 324
3 500
Msw+pl 1 2580 5 Normal
(kNm)
2 118
3 598
MLL (kNm) 1 966 10 Gumbel
2 804
3 1099
1=support of arch 1, 2= quarter of arch 2, 3= midspan of arch 2
Table 13.- Statistical definition of random variables
The actions due to the live load were obtained from a probabilistic model for traffic flow simu-
lation that uses the influence lines of the effect under study [12]. The values of the reliability
index obtained for the sections at support, quarter-span and mid-span are 1 = 1.65(Pf =
0.49x10-1), 2 = 4.82 (Pf = 0.79x10-6), 3 = 1.14 (Pf = 0.127). From these low values of the
reliability indices it is clear that the bridge safety is really low for the existing traffic on the
bridge. However, the bridge is carrying the actual loads perfectly. The conclusion was then
that the model does not accurately reflect the real behaviour. For this reason, the system
reliability analysis was performed.
6.1.3 Assessment based on non-linear response and system failure analysis

As explained in 4.1.3, it is assumed in this case that, provided the ultimate strength of the
material is not reached at any section, the bridge fails when a sufficient number of plastic
hinges (normally 4) develop in different cross-sections. In this case, the Limit State Function
can be written as:
G= R-S (21)
R is the ultimate resistance of the bridge taken as a system, i.e., the load that the bridge can
carry up to the development of the failure mechanism. S is the external traffic action. In order
to statistically characterize the random variable R, a simulation process was carried out using
a deterministic model for analysis of masonry structures up to failure [4]
Description of the non-linear numerical model
The numerical model performs the nonlinear analysis of masonry spatial structures consist-
ing of curved, three-dimensional members with variable cross section. The detailed deriva-
tion leading to the global equations of the structural problem, together with some details re-
garding its numerical implementation, may be found in the paper by Molins and Roca [4].
The application of the model to multi-span arch bridges is demonstrated in [19].
The global constitutive material model adopted results as a combination of partial constitutive
equations for masonry subjected to tension, compression and shear stresses. Interaction
between the normal and shear responses is considered through the use of a biaxial strength
envelope in order to define the limiting set of strength values.
The behaviour of the material subject to compressive stresses is described by means of a
bilinear elasto-plastic model. While in tension, masonry is modelled as a simple linear elastic
perfectly brittle material. Either cracking or separation between blocks occurs once the ten-
sion strength of the fabric is reached. No further softening behaviour is considered after the
first cracking. Shear behaviour is modelled by means of a bilinear elasto-plastic curve, similar
to that adopted for brick masonry under compression, with the limit strength max determined
from the strength envelope. In this case, a Mohr-Coulomb criterion, with no cohesion, was
adopted in order to model the shear response as a residual resisting mechanism due to fric-
tion between blocks and mortar in joints. The shear strength max is thus expressed as:
max = (22)
where is the applied compressive strength and is a coefficient of friction. Under tension,
the shear strength is made equal to zero.
As has been demonstrated in chapter 4.3.2. realistically evaluating the load capacity of barrel
vault bridges, such as Magarola, requires taking into account the horizontal action of the infill
and the spandrel walls, in addition to its dead load.
Figure 83.- Modelling of an arch bridge with the infill as a set of equivalent linear members.
An approach based on its discretisation into a system of equivalent linear elements is

adopted. This equivalent system is composed of, firstly, a series of curved elements to model
the arch ring, plus a second series of tapered members with axes parallel to the platform of
the bridge, to model the spandrel walls and infill. A third series of joint elements are then
used to link the spandrel and infill elements to the ring. These joint elements are defined ab-
solutely rigid in the direction normal to the extrados, while flexible in the direction tangent to
it. Relative rotations between arch and infill sections are allowed (Figure 83). The validity of
such a technique of modelling has been assessed by detailed comparison with results ob-
tained using a conventional FEM plane stress analysis. Satisfactory engineering agreement
was observed for both elastic linear and non-linear analyses [20]. Using the presented tech-
nique of modelling each arch ring is discretized into eight elements, and eight infill elements
are connected to the ring elements by joint elements. Piers are modelled using one element
of variable cross-section. Connections between the upper node of the piers and the springing
node of the arches are modelled by rigid links. The whole model contains 131 elements, in-
cluding the joint elements.
Additional hypothesis used in the analysis are: (1) the axles distribute the load uniformly in
the cross section, (2) the favourable effect of the longitudinal distribution of the load through
the infill is neglected, and (3) abutments are considered absolutely rigid. The second hy-
pothesis is reasonable when working with vehicles of various axles which, in fact, produce a
more distributed load pattern.
In all the analyses performed, dead load is first applied in one step. Successive increments
of live load are then applied until reaching failure. The failure is due to the formation of a 4-
hinge mechanism ( see figure 84).
Figure 84: Stress state of Magarola bridge at failure.
Non-linear analysis of the structure under dead loads showed that the arch rings are quite
uniformly compressed with the exception of a short zone at the extrados of the springings.
Calculation of traffic action

The first step in the calculations was to determine the most unfavourable type of vehicle and
its worst position over the structure. It was difficult to define a priori which one of the vehicles
representative of heavy trucks in Spanish highways would be worst because the back bogie
of the three axle vehicle concentrates more load than the bogie of the five axle vehicle (three
axles in the bogie) which is the heaviest. It was found that the five axle vehicle acting very
close to mid span of the second arch was the most unfavourable for this bridge. However,
the differences between applying the load on the second arch or on the third are slight.
Figure 85.- Load factor vs. position for Magarola bridge.
Figure 85 shows the evolution of the load factor () for a five axle vehicle along the second
arch. This factor means the number that applied to the axle loads of the average 5-axle truck
(400 kN) produces numerical prediction of failure. In all positions failure is due to crushing in
compression at the crown of the second arch. That crushing occurs for a quasi collapsing
mechanism involving the formation of three hinges in the arch, one in each pier (not com-
pleted), three in the first arch of the bridge and two (not completed) in the third arch (adja-
cent to the second). Figure 86 shows the deformed shape of the bridge at failure. As can be
seen, the contour conditions induced by the flexibility of the piers and the adjacent arches
increases the flexibility of the loaded arch. They also permit detecting the huge rotation ex-
perienced by the sections around the crown of the second arch. In fact, the comparison of
these results with those obtained supposing rigid supports at the springings of the arch
shows that the loss of load capacity induced by the actual support conditions is around 50%.
Performing a simulation process based on the Uptaded Latin Hypercube Method [6] and us-
ing a Kolmogorov-Smirnof fitting technique, it was found that the load factor,, is well ap-
proximated by a Lognormal variable with mean equal to 56.7 and standard deviation 10.6.
The surveyed traffic data with the distribution of total weight in 5-axle trucks, identified S as a
Normal random variable with mean 2.0 (two 5 axle truck side by side) and a coefficient of
variation (COV) equal to 25 %. According to these values, a reliability index = 12.9 was
obtained for the actual traffic. As can be seen, the safety of the bridge using a reliability-
based methodology with a suitable and accurate model of the bridge behaviour up to failure
can be assured. Much higher reliability level is obtained compared to the linear analysis.
In conclusion, when a model that fully reflects the complex and non-linear behaviour of multi-
span arch bridges is used, the corresponding reliability indices are high enough to assure the
correct performance. Contrary, the assessment has also shown how assuming a simplified
and not as accurate model, the corresponding results may be completely unrealistic and lead
to inappropriate decisions concerning bridge safety. The non-linear model predicts a mecha-
nism of collapse involving adjacent piers and arches. The analyses carried out showed the
importance of taking into account the complete structure instead of a single arch. In fact, the
analysis for single arches would have produced an overestimation of 100% of the load ca-
pacity. These two conclusions are very important regarding the future assessment of other
multi-span arch bridges
Figure 86.- Deformed shape at failure of arch 2, Magarola bridge.
6.2 Assessment of Fatigue LS
The assessment of the fatigue performance of a bridge can be formulated in two different
ways:
1.- Calculation of the reliability index or probability of failure in a given reference time
2.- Calculation of the remaining service life with a predefined probability level.
With the available experimental data, the first assessment will be only pertinent in the case of
fatigue due to compression in the masonry (single-ring arches). The process of the calcula-
tion of the reliability index or probability of failure due to service loads causing a fatigue fail-
ure is as follows:
1.- Build up a theoretical model of the bridge. Because the load level introduced in the analy-
sis will be equivalent to the service loads, a linear elastic model or other simplified models as
presented in this document can be used.
2.- Define the typical trains passing the bridge determining the axle-spacing and axle-loads.
Define how many of each typical trains will cross the bridge in the remaining service life pe-
riod (defined by the user)
3.- Simulate the passage of each train in the numerical model and obtain the stress versus
time relationship for the most critical locations in the bridge. Take into account that the loads
should include the dynamic amplification factor as we are dealing with service load levels.
4.- From the stress-time relationship obtain the maximum stress in the predefined locations
and obtain the relation S= maximum stress/ compressive strength for each typical train and
only for the stress peaks in the stress-time curve where S > 0.5 (endurance limit)
5.- For each value of Si > 0.5 and according to the number of trains that will cross the bridge
in the defined period of time, obtain the number of times, ni , that each value S will be
reached in the selected locations ( Define the histogram of S ). According to what is ex-
plained in the next point, it is recommended to obtain the values of ni for the ranges
0.5<S<0.55, 0.55<S<0.6, 0.6<S< 0.65, 0.65<S<0.7, 0.7<S<0.75,0.75<S<0.8, 0.8<S.
6.- According to the Miners rule, the limit state function can be defined as:
ni
G=1
S i > 0.5 N i
(23)
where Ni is the number of loading events causing the failure. According to what is explained
in this document, Ni is a random variable that can be modelled by a Weibull distribution with
the parameters as presented in table 14 depending on the value of S. For each of the 7
ranges of S defined in point 5, the corresponding parameters of the distribution can be ob-
tained from table 14.
S lnu u
> 0.8 5.336 528
0.75-0.8 4.3585 3436
0.70-0.75 7.7298 1324
0.65-0.70 4.8822 40306
0.60-0.65 2.521 40010
0.55-0.60 5.3679 353144
0.5 - 0.55 9.0308 29138
Table 14.- Parameters of the Weibull distribution as a function of the stress level
7.- Once the random variables Ni present in the limit state function are perfectly defined, then
applying FORM, the reliability index can be calculated.
The process of the calculation of the remaining service life with a predefined probability level
is as follows:
1.- Proceed from step 1 to five as in the previous case
2.- Calculate the minimum stress level (compression or shear) in the selected locations. This
will be the value corresponding to the empty bridge (self-weight + permanent loads). For
each stress range, calculate the ration Ri = minimum stress/maximum stress
3.- In the case of fatigue due to compression (single-ring arches), according to equation 17
and the values of A and B in table 10 for the desired probability level, calculate for each
range of S (with the corresponding values of R) the number of events to failure Ni
4.- In the case of fatigue due to shear (multi-ring arches), at this moment the only fatigue
equation available is equation (20), which corresponds to a probability of 50 % (correlated
directly to test data). Using this equation, the number of events to failure Ni can be estimated.
5.- Given the estimated number of typical trains crossing the bridge for a reference period ( 1
year for instance), then calculate the time necessary to reach the value 1 in the accumulated
damage equation.
7 Conclusions
As used for other materials like concrete and steel, the potentiallity of probabilistic methods
in the assessment of masonry arches has been studied. The main problems encountered for
such an approach are mainly the response behaviour of this type of bridges that makes very
difficult an accurate modelization as well as the different construction methods used since
many years (single-ring, multi-ring,.). Many different materials as brick, mortar, infill mate-
rial with very different strength properties also introduce challenging problems to the model-
ling. Finally, and of special focus in the case of a probabilistic approach, is the lack of ex-
perimental data available in the literature necessary to derive a statistical definition of the
main properties of the materials. Since an accurate probabilistic assessment mainly relies on
an appropriate model of the structure to simulate the real behaviour and on an appropriate
statistical definition of the random variables involved, still an important work has to be devel-
oped in the coming years for a correct estimation of the safety of these bridges in terms of
their probability of failure or reliability index. However, with the still limited experimental data
available, some experiences have shown the potentiality of probabilistic methods applied to
masonry arches, resulting in final assessments that may leave the bridge as it is without
costly repair and/or strengthening remedial measures.
The present document is a first step in this sense, as it presents a complete methodology for
the probabilistic assessment of masonry arches at the serviceability and ultimate limit states.
The document explains the definition of the different failure modes and corresponding limit
state functions that may occur depending on the type of masonry construction (mainly single-
ring and multi-ring). Also the introduction of the possibility of the fatigue failure of masonry
arch bridges and the proposal of new assessment methods based at the serviceability level
[21] and not only at the ultimate level of the 4 hinge-mechanism is a promising initiative in the
field. As a consequence, the present document has deeply investigated the potentiality of a
probabilistic approach in the estimation of fatigue safety and remaining service life of ma-
sonry arch bridges. As a result, the need of simple numerical models to perform fast simula-
tion trials has been enhanced. Thus, the influence of important features in the bridge re-
sponse, as the arch-backfill interaction, the modelling of multi-ring arches and the bridge
skewness in the development of simplified theoretical models at the ultimate and serviceabil-
ity limit states have been worked out too.
The analysis developed in the present work has verified the results obtained in laboratory
tests regarding the influence of skewness in the failure load by 4-hinge mechanism. In fact, it
has been confirmed that the influence of a skew angle less than 22.5 is almost negligible.
The study has confirmed that only with advanced 3D FEM models is possible to correctly
model the behaviour up to failure of masonry arches with skew bigger than 22.5. It has been
also concluded that the worst load position for the arch can be deduced using a simple
model.
In summary, a methodology is presented for the fatigue and serviceability assessment of
masonry arches, and probabilistic S-N curves for the behaviour of masonry under fatigue are
developed based on the few experimental data available. In the case of masonry under com-
pression, a fatigue equation with various levels of probability of failure is also proposed that
may be used for deterministic assessments. In the case of masonry under shear a fatigue
equation with a 50 % confidence level is also proposed. The last can be of relevant applica-
tion to the case of multi-ring arches. Of course, these are only preliminary models and these
fatigue models have to be updated as more experimental data becomes available from future
laboratory and full-scale tests.
8 References
[1] D4.7.1. Structural assessment of masonry arch bridges. Background document. Sustain-
able Bridges, 2007.
[2] Casas, J.R.: Assessment of masonry arch bridges. Experiences from recent case stud-
ies. Current and future trends in bridge design, construction and maintenance. Thomas Tel-
ford, 1999, pp. 575-586
[3] Casas, J.R.; Molins, C.: Assessment of the Magarola arch bridge. A case study. Pro-
ceedings of the Sixth nternational Conference on Structural Studies, Repairs and Mainte-
nance of Historical Buildings, Dresden, 1999, pp. 333-342
[4] Molins, C.; Roca, P.: Capacity of masonry arches and spatial structures. Journal of
Structural Engineering, ASCE, 124, (6), 1998, pp. 653-663
[5] Page, J.: Masonry arch bridges. State of the art review. Transport Research Laboratory,
UK, 1993
[6] D4.4.3. Probabilistic non-linear analysis. Background document. Sustainable Bridges
2007.
[7] Martn-Caro, J.A.; Martnez, J.L.; Len, J.: A first level structural analysis tool for the
Spanish railways masonry arch bridges. Proceedings of Arch Bridge IV, Advances in As-
sessment, Structural design and construction. CIMNE, Barcelona, 2004, pp.192-201
[8] D4.7.3. Methods of analysis of damaged masonry arch bridges. Background document.
Sustainable Bridges, 2007.
[9] Roberts, T.M.; Hughes, T.G.; Dandamudi, V.R.; Bell, B.: Quasi-static and high cycle fa-
tigue strength of brick masonry. Construction and Building Materials, Vol. 20, 2006, pp. 603-
614
[10] Melbourne, C.; Tomor, A.; Wang, J.: Cyclic load capacity and endurance limit of multi-
ring masonry arches. Proceedings of Arch Bridge IV, Advances in Assessment, Structural
design and construction. CIMNE, Barcelona, 2004, pp. 375-384
[11] Crespo-Minguillon, C.; Casas, J.R.: Fatigue reliability analysis of prestressed concrete
briges. Journal of Structural Engineering, ASCE, 124, (12), 1998, pp. 1458-1466
[12] Crespo, C. ; Casas, J.R., A comprehensive traffic load model for bridge safety checking,
Structural Safety, 19, (4), 1997, pp. 339-359.
[13] ASCE: "Foreword to Fatigue and Fracture Reliability. A state of the Art Review ". Jour-
nal of the Structural Division,ASCE, 108,1982, pp. 1-88.
[14] Hwan, B.: " Fatigue analysis of Plain concrete in Flexure ". Journal of Structural Engi-
neering, ASCE, 112, (2), 1986, pp. 273-289.
[15] Shi, X.P.; Fwa, T.F., Tan, S.A.: Flexural fatigue strength of plain concrete. ACI Materi-
als Journal, 90, (5), 1993, pp. 435-440
[16] Castillo, E. and Fernndez, A.: "Statistical models for fatigue analysis of long elements".
Proceedings of the IABSE Workshop on Length Effect on Fatigue of Wires and Strands, El
Paular, Madrid, 1992, pp. 15-32.
[17] Fernndez-Canteli, A., Castillo, E., Argelles, A.: "Length effect on fatigue of wires and
prestressing steels". Proceedings of the IABSE Workshop on Length Effect on Fatigue of
Wires and Strands, El Paular, Madrid, 1992, pp. 125-135.
[18] Gilbert, M.; Melbourne, C.: Analysis of multi-ring brickwork arch bridges. Arch Bridges.
Proceedings of the First International Conference on Arch Bridges. Thomas Telford, 1995,
pp. 225-238
[19] Molins, C, Numerical simulation of the ultimate response of arch bridges. Chapter of the
book Structural analysis of historical constructions II, ed. Roca, Gonzlez, Oate and
Loureno, CIMNE, Barcelona, 1998, pp. 93-123
[20] Sicilia, C., Ultimate analysis of masonry bridges (in Spanish). Master Thesis. Universitat
Politcnica de Catalunya (UPC), Barcelona, 1997.
[21] Melbourne, C; Tomor, A. : A new assessment method for masonry arch bridges. Pro-
ceedings of IABMAS06. Taylor and Francis. Porto, 2006.
[22] Melbourne C., Hodgson J.A.: The behaviour of skewed brickwork arch bridges. First
International Conference on Arch Bridges, Thomas Telford. London, 1995, pp. 302-320.

Masonary Arch Bridges

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Masonary Arch Bridges

Caricato da

Copyright:

Formati disponibili

Masonry Arch Bridges

Background document D4.7

Copyright Authors 2007.

Project acronym: Sustainable Bridges

Project co-funded by the European Commission

D4.7.1 Structural assessment of masonry arch bridges

Part D4.7.3 is focussed on the effect of damages on carrying capacity.

INITIAL INTERMEDIATE ENHANCED

INITIAL INTERMEDIATE ENHANCED

Part Responsible Contributors Reviewer

Copyright Authors 2007.

Project acronym: Sustainable Bridges

Project co-funded by the European Commission

2.4.3 Finite element models....................................................................... 60

Fatigue performance of multi-ring brickwork arches

Soil-structure interaction in masonry arch bridges

1. Masonry arch behaviour under long-term cyclic loading

a) Four hinge mechanism c) Sliding

b) Ring separation d) Abutment movement

a) Single-ring c)multi-ring b) multi-ring with headers

1.3 Arch construction

1.3.1 Arch dimensions

Table 1-2. Max. number of bricks in arch rings

1.3.2 Material properties

Table 1-3. Material properties

1.3.3 Loading arrangement

Figure 1.3: Loading frame

Figure 1.4: Hydraulic jacks

Figure 1.5: Deflection gauge positions

1.4 Static loading tests

1.4.1 Strong brick arches

1.4.1.1 Arch A (3m span)

Figure 1.6: Arch A: failure mechanism

Figure 1.7: Arch A: Failure mechanism

Figure 1.8: Arch A: Load-Deflection

1.4.1.2 Arch G (3m span)

Figure 1.9: Arch G: Failure mechanism

Figure 1.10: Arch G: Failure mechanism

Deflections, measured at and span show slight change in curvature around

Figure 1.11: Arch G: Load-Deflection

Figure 1.12: Arch G: Load-Strain

Load (kN) Load (kN)

1.4.1.3 Arch M (5m span)

Figure 1.15: Arch M: Crack locations and AE positions

Figure 1.16: Arch M: Failure mechanism

Figure 1.17: Arch M: Ring separation failure

Figure 1.18: Arch M: Collapse mechanism

potential areas of damage.

Load (kN) Load (kN)

Figure 1.19: Arch M: AE amplitude vs. Load

1.4.2 Weak brick arches

1.4.2.1 Arch T (5m span)

Figure 1.20: Arch T: Failure mechanism

Figure 1.21: Arch T: Failure mechanism

Figure 22: Arch T Amplitude vs. load

1.5 Cyclic loading tests

1.5.1 Strong brick arches

1.5.1.1 Arch C (3m span)

Figure 1.23: Arch C Failure mechanism

Figure 1.24: Arch C: Ring separation

Figure 1.25: Arch C: Sliding failure

Figure 1.26: Arch C: Collapse mechanism

1.5.1.2 Arch E (3m span)

Figure 1.27: Arch E: Failure mechanism